445 texbook 0003

290 OPTIMAL DATABASE MARKETING

Examining the rest panel where the new format and copy were changedsimultaneously, you find no difference in response versus the control package(3.47% vs. 3.50%). On the basis of this result, you would discard bothasan option and try again. However, when you examine the test resultsofeachelement tested separately) notice that the new copy approach was a winner.Without these separate element test panels, this would not be evident.Remember, testing is the foundation upon which you will grow your

company. Test wisely.

Rule 4: Test for Only MeaningfulPackage Element Interactions

Generally, it is unnecessary to test every possible package element comhin-ation in your test plan. For example, the product manager at ACMEDirect may be interested in testing the following changes to the comrolpackage:

• Price increase• Addition of a premium• Addition of an action device• New format

Testing every possible combination of price, premium, format, andaction device yields a total of 16 test panels. Testing all 16 panels iscalleda full factorial test design.The only reason a marketer would test a full facrorial test designwouldbe

if it was believed that interactions will occur between all four elementswithrespect to response. In this example, the only possible interaction to becon-cerned with would be between price and premium. In other words, if restinga higher price, perhaps the minus in response (due solely to pricing]wouldheless for the package with a premium versus the package without the premium.Note that other interactions may also be possible, bur you wouldneed

to know more details about the other panels ro make a proper judgmenrcall. For example, if the action device is a scratch-off card in whicheveryone gets a $1 disco t" . . h .un , It too may mreracr wit pnce.Assuming ACME 0"· " I' " " o~ihl1'. irect IS on y interested In assessing a P

tnteractrrm betwe . d" . Id ar as. en price an premium the test senes wou appeshown III Exhibit 14.2. 'On the basis of this test series, how will ACME Direct determul' ifthe

addition of a prerni th f b ne!". rum to e control package offset any or all 0 rne ..trve effect that a . . i rr Dire<'ill d . pnce lllcrease might have on response? AC1V.u..

W etermme this bye" I" P 1'1 and2. xanunmg t re index in response for Test ane5versus the d .111 ex III response for Test Panels 3 and 6.

---------------------Plwningand Designing Marketing Tests 291

Exhibit 14.2 ExampleTest Series

Test Panel Description

I2J456

Control test packagePrice test package-as control package with $2 price increasePremium test package-as control but with premium for order addedFormat test package--as control but with a new formatAction device test package-as control but with an action device addedPrice and premium test package-as test panel #3 with $2 price increase

Onedrawback to nor testing every possible element combination is thatj'ouare left with forecasting issues in rollout. For example, considering thersrseriesshown in Exhibit 14.2, what would happen if ACME Directdcridesto roll out with the addition of both the premium and action device10 thecontrolpackage for their next major marketing campaign? How willtheylorecastthe lift in response they expect in rollout, given that they didnmtest this combination? To determine this, most major direct marketerssimplyadd all individual element pluses and minuses and take a certainpercentageof them for the final response forecast. The percentage taken is~picallyderived from historical information or experience. It will vary'omdirectmarketer to direct marketer and be based on the number and~pesofelements being added together. In orher words, if the addition ofthepremium gave you a +25% in response versus the control and theaddirionof the action device gave you a + 100/0 in response versus thecontrolldo not expect to get a +350/0 in response if you incorporate boththingsinthenew package simultaneously. You will get something less thanthat.Theywill not be additive.

Rule5: Define the Universe for Testing Carefully

Catefulconsideration must be given to the names selected for testing.Mal'or keri 'I d th core or primarymar enng elements are typical y teste to e . .CUstomersegment as a whole. They should not be tested to selects wjthintheco ' . k ft made by small. re Or pnrnary customer segment-a rmsra e a en .direct k ." onsistency 111 testmar erers. Why ISthis an Issue? Because you want ctesultsY . to another. If the. au want to bridge test results from one test sertessame d f . . . cannot accurateJyf e Inltlon of names are not tested over time, you IOtetas,lifr f Its For examp e,. I Of future campaigns based on past test restr . , .,nuwa f this years campmgn.II nr to use the action device tested last yea r or 1

unforru If' if t1y from those you, natey, the names tested last year dif er srgru ican Iif '~111be I forecast the I t JJ1promoting this year. You cannot accurate y I Iresponf Its to a comp ere yd' se Or one Customer group based on test resuIffetentCUStomer group.


If you are testing a custom format designed for a unique customer segmen~you must also include within your test plan a rest of the same uniquecustomersegment to receive the control format. For example, in the case of resting anewformat designed specifically for converting non buyers, you will of coursebetesting it to the segment of non buyers and not your core customer segment.Togauge the success of this new non buyer format, you must make sure you havea panel within your test plan of nonbuyers receiving the standard controlpackage. You will not be able to compare the resuJrs of this custom formarsento the nonbuyers to the control package sent to the core cusromer segment.Similar to comparing apples to oranges, this will yield misleading results,

Outside List Test Design Considerations

Regarding outside list testing, you may be well advised to test for interactionsbetween lists and the package elements (format, price, offer, etc.). In orherwords, certain lists, depending on how they are sourced, may react differentlyto different package elements. For example, a list sourced via a sweepstakespromotion will in all likelihood respond berter to a sweepstakes promotionthan a nonsweepstakes promotion. A list sourced via a soft free trial offerwiUin all likelihood respond better to a similar offer than to a hard cash-with-order offer. If you do not see any reason to test for list by packageelementinteractions, don't. Use yOUT experience to guide you in setting up thetest.

For example, if you are testing five lists, two formats, and twooffers}a full factorial test design will require that you test 20 different combinatineof list, format, and offer. If you have a fixed testing budger, you mayno!beable to test enough names within each ceU to get meaningful results.Inorherwords, the confidence intervals constructed around the response rateassociated with each cell may be so wide that you cannot make a decision(see Chapter 13). If this is the case, we advise you to reduce the numberoftest cells, keeping only those interactions that make me most sense.Knowledgeable lisr brokers should be able to help you. On the basisofrheuexperience, list brokers should know what types of offers work bestfurwhich lists. Take advantage of their expertise.

For example, ACME Direct is going to test five lists, two formats,a.ndtwo offers, all within a budget of 25 000 total names. A fuUfacrorialdesignwill yield 20 test II - h 1 2 ' , E hibu14.3-. ce s Wit , 50 names per cell, as shown III .x I I

W~th 1,250 names per cell, the test results will have so much error varianceassoctared with them that they will be difficult to interpret. Exhibir14.4shows the 950/. nfid b b sedono co I ence ounds around various response rates asample sizes of 1250 ' f h ·nrerv,1, names. As you will notice the Widths 0 eac Iare so wide that 'you cannot make a sound decision.

Another issue arises regarding this test design for the "offer" resrpanels,TYPically for off h . Fo!such, er tests, t e payment rates are also of Interest.

------------- ...-Planningand Designing Marketing Tests 293

Exhibit14.3 FullFactorial Test Design

Format I Format 2

lilt Offer 1

~IA 1,250IiltB 1.250IiltC 1.250IistD 1,250li1E 1.250

Offer 2 Offer I Offer 2

1,2501,2501,2501,2501,250

1,2501,2501.2501,2501,250

1.2501,2501,2501.2501,250

Exhibit14.4 Example 95% Confidence Bounds Based on 1,250 Names

Ilrsponse Rates ofrett on 1,250Names (%)

95% Confidence BoundsAround the Test Response Rate (%)

1.52.03.04.05.06.0

0.83 to 2.171.22 to 2.782.05 to 3.952.91 to 5.093.79 to 6.214.68 to 7.32

small samplesizes, few orders will fall into each cell, implying an evenlarger variance associated with the payment rates.

If increasingthe sample sizes of the rest cells is not an option, ACMEDireerneedsto eliminate any cells that are not necessary. For example, ifthereis no reason to believe that "list by package element" interactionsexist,ACME Direct will not test a full factorial design. Instead, they willI'ltthe format on one list only and the offer on one lisr only. This allows~emto testmore names per cell and increase the precision of each cell's"ponserate while sraying wirhin their testing budget (see Exhibit 14.5).IfACME Direct sees no reason ro believe there will be an interacnon

~nveenformat and offer, they can further reduce the number of test cellsblonemore,as shown in Exhibit 14,6.

~t 14.5 Reduced Factorial Test Design

Formot IFormot 2

list Offer 2_________ -.:O~ffi~er'___/~ _=0'!iffir.=e:._r=-2 ~O~ffi~e~r ~1-------;:-m-U'A 3.125 3./25u"B 3,125 3.125~tC 3.125UltD 3.125

~tE 3~.~12:5 -------------____ 3,125

Offe, 1 Offer I


Exhibit 14.6 Further Reduced Factorial Test Design

Format I Format 2

Ust Offe, 1 Offe, 2

List A 3.571 3.571List B 3.571List C 3.571List D 3.571List E 3.571

3.571

On the basis of the final test design shown in Exhibit 14.6, how\\illACME Direct determine the winning format? They will compare "Format1 & Offer 1" versus "Format 2 & Offer 1" on List A.How will ACME Direct determine the winning offer? They will compare

"Offer 1 & Format 1" versus "Offer 2 & Format 1" on Lisr A.How will ACME Direct forecast the response rate for List C if promoted

witb Format 1 and Offer 2? They will determine the lift in responseobserved for List A when given "Format 1 & Offer 2" versus "Format 1&Offer 1." Once determined, they will use it to index up the responserateobtained for List C (which was based on "Format 1 & Offer 1").

Sample Size Considerations

With the test panels determined, the next most important parrof theplanning process is to determine the appropriate sample sizes for each.Without adequate sample sizes, the time and money spent creatinganddeveloping your marketing tests will be wasted. This is true whetheryouare planning direct mail format Or telemarketing script tests.

You have spent the time and money to create and develop new restpanels, so why risk misinterpreting the results of the tests by testingfewernames to save a few dollars? Without proper sample sizes, your testresultswill have variance so great that the results will be unreliable. As surh,odds will be that the test results will not resemble what you can expecrifyou decide to roll Out with the new test promotion to a larger audience.In other words, the range in which the actual test result could lie (yourconfidence interval) will be so wide that it will be virtually meaninglessand therefore difficult to make a decision with any certainty, as was seenin Exhibit 14.4.It is . . fewImpOrtant to understand the ramifications of promonng roo

names. Small sample sizes can result In two possible scenarios that couldnegatively impact your company:

---------------------Planningand Designing Marketing Tests 295

I. You misread the test result and believe tile new test panel hasbeatenyour current control package, when in fact it will do worse.

1. Youbelieve your current control package has beaten the new testpanel,which was in fact a winner.

Bothsituations can negatively affect the long-term profit of yourrornpany.Weigh the savings of promoting fewer names against the costs ofIosrrevenueor lost opportunity caused by misreading test panel results.Withtrue assessment of the costs and savings, you will undoubtedlyronc!udethat it is a wiser business decision to test more names. If yourliSting budgetdoes not allow you to test more names, cut back on testpanels,nor the number of names tested per panel.Wenextdiscuss three applications for sample size determination. In par-tklllar,we show you how to determine the appropriate sample sizes wheninterest revolves around the accuracy of a single sample mean, a singlesampleproportion, and the difference between two sample proportions.

Sample SizeDetermination for a Sample Mean

Wheninterest revolves around ensuring that a sample mean will be withinarmain range of the true population mean, you will employ the formulaprovidedin this section to determine the required sample size. For example,you are interested in closely approximating the mean dollars spent perCUltomerorder for a new catalog promotion based on a test.

If interest revolves around determining the required sample size whentoncernedwith the reliability of a sample mean, yOll need the followingmformation:

• Themaximum allowable error variance EThis measure represents the maximum allowable error you arelull·· . F pie you desire theI mgto accept 111 your sample estimate. or exam ,sa"'plemean to be within :<:$1.50 of the true population mean. Inthisexample, $1.50 is the allowable error variance.

I Anestimate of the sample standard deviation 5 . .o be basi . th auerage deVIatIOnn t ie aS1Sof past data, you need to estimate e e»a d h ·11 b b d in your sample.rO/ll1 t e sample mean you believe un e 0 serveIfill doubt, it is best to err on the high side.

I Thedesired error rate "'% and confidence level (1 - "')% . hIf, t. . t of $81.50 Wit'lor example, your sample yields a mean est/ma e. I flall /1 b 50 950' confidence leve WIa DLVa Ie error variance set at $1. , a /0 f' '·11 . I de the truegllarallteethat the interval $80.00 -$83.00 WI me u h IIP . .. w: mffldt~a0pu/atlon mean with 95% probability- e recorn IidSa / . 900/, . better conti encernpe Size calculations be perf01·med at a ° 01

296 OI'TLMAL DATABASE MARKETING

level. Employing lower levels of confidence drastically increasestiJerisle (error rate a) of obtaining misleading results, which could leadtoincorrect and costly decisions.

Once you have all this information, the required sample size is calcularedas

n=[(Z2)(S2)]

E2

WhereS is the estimated standard deviation associated with the sample measureE is the maximum error you are willing to acceptZ is equal to 1.645, 1.96, and 2.575 for 90%, 95%, and 99% confidence

levels, respectively (z values associated with other confidence levelscanbe found in Exhibit 13.1).

Sample size formulas are derived from the same formulas for confidenceintervals.

Note that if the actual sample standard deviation achieved is greaterthan estimated, the error variance around the test estimate will begreaterthan what was specified as the maximum allowable error variance. (Intheother hand, if the actual sample standard deviation achieved is lessthanestimated, the error variance around the test estimate will be lessthanwhat was specified. If in doubt about your estimate of the sample standarddeviation, it is best to err on the high side. This will safeguard againsttheresulting variance being greater than your desired maximum allowablevarrance.

To illustrate further, the market research director at ACME Directisprepared to survey a sample of active customers on the database in anattempt to estimate their true annual direct marketing expenditures fromallsources (catalog, lnternet, direct mail, telemarketing, infomercials, etc}andall companies. An accurate estimate is required. In fact, he wants theestimate of annual expenditures to be within $5.00 of the true mean forallactive Customers with 950/0 confidence. First, he has to estimate thestandard deviation of expenditures. He has no data yet, so there is nowa.y

to determine this value except by estimation. The easiest way to do thisISto estimate the likely mean and the likely minimum and maximum valuesthat would account for roughly 95% of all active customers on the ACMEDirect database. Once estimated, he will take the absolute value of thelargest diHerence between the estimated mean and minimum valueandestimated mean and maximum value and divide by 2. This will giveanesttmate of the standard deviation in the data likely to be observed.

For example, if he estimates the mean to be approximately $100.00andthe mmunurn and maximum values to be $25.00 and $200.00, respecnve!y,then the absolute differences between the mean and rhe minimum and

lII!a~-----------------__Manningand Designing Marketing Tests 297

~aximumvaluesare $75.00 and $100.00, respectively. The largest of thesefl'iO differencesIS $100.00. DIVIdIng this figure by'2 yields an estimated;taJ1darddeviation of $50.00.

In thisexample,

E = $5.00S ~ $50.00 (as just determined)z = 2.575 for a 99% confidence level

Pluggingall known data into the formula provided yields the followingsamplesize estimate:

n[(Z2)(S2)]

£'[(1.96')(50')]

5'[(3.8416)(2,500)]

25

n=

n=

9,60425

n = 384.16

n=

The market research director needs to obtain information from 385CUstomers in order to guarantee with 950/0 confidence that his estimate ofd' , "llfel marketing expenditures will be within :':$5.00 of the true meanexpetlditures.Assumethe survey is complete and the market research director

sO'Vel'ed385 k i hi. customers. Also assume that he was right on the mar LD 15

~lImateof rhe sample standard deviation at $50.00. If he now constructsaroofide' I d he inf .. neemterva around the sample mean base on t e In ormationprOVidedin Chapter 13, how wide will the interval be? Under theseOrcllmstances,the width on either side of the mean should be :,:$5.00,lUll ashedesired.~Howwould the resulting confidence interval be affected if the actualoorvedstandard deviation of the sample is greater than the estimated\,1" of$'0 '11b 'd th, . ) .OO?The resulring 95% confidence interval WI e WI er an-hoo ' h 'Oneit er Side of the mean estimate.~ow would the resulting confidence interval be affected if the actualOf~ed standard deviation of rhe sample is less than the estimared valueII$JO,OO?The resulting 95% confidence interval will be less rhan :':$5.00"'either id fT SI eo the mean estimate. .t odeterminethe appropriate confidence level, the market research direc-IIrwI11g h 3 f d mining thea a t rough the steps outlined in Chapter 1 or eter I~' h'pnateconfidence level for confidence intervals and hypot eSIS rests.

298 OPTLMAL DATABASE MARKETING

Sample Size Determination for a Sample Proportion

When you are interested in ensuring that a sample proportion will bewithina certain range of the true population proportion, employ the formula pro-vided in this section to determine the required sample size. For example,you are interested in closely approximating the response rare for a newproduct offering or new list of prospects based on a test.To calculate the sample size required when concerned with the reliability

of a sample proportion, you need the following information:

• The maximum allowable error variance EThis measure represents the maximum allowable error you arewillingto accept in your sample estimate. For example, the test responseratemust be toithin tz 0.0025 (0.25%) of the true response rate that willbe achieved in rollout. In this example, 0.0025 is the allowable errorvariance .

• An estimate of the sample proportion pEstimate the response rate (sample proportion) you expect to receivein the test. This can be based on prior information or a bestguess.

• The desired error rate a% and confidence level (1 - «)%If, for example, a test yields a 3.00% response rate (as expected) witban allowable variance set at 0.25%, a 95% confidence level willguarantee that the interval 2.75%-3.25% will include the truepopulation response rate with 95% probability. We recommend thatall sample size calculations be performed at a 90% or bette:confidence level. Employing lower levels of confidence drasticallyincreases the risk (error rate ex)of obtaining misleading results,whichcould lead to incorrect and costly decisions.

Once a ll the information is known, the required sample size is calcuJaredas

11.=[(z2)(P)(1 - p)]

£2

Where

p is the estimated sample proportionE .is the maximum error you are willing to acceptZ IS equal to 1.645, 1.96, and 2.575 for 90%, 95%, and 99% confidencelevels, respectively (z values associated with other confidence levels canbefound in the Appendix).

-.;....._-----

. .Note that if you are unsure of the response rate you will achieve in resr,It is best to err on the side that places the proportion closer to 0.50. In other

..-------------------~J/1ningandDesigningMarketing Tests 299

words,if you believe the sample response rate will be somewhere between15%and 4%, use 4% as your estimate, because it is closer to 50%. If you"lieverh'sampleresponse rate will be somewhere between 82 % and 90%,[Be 82 becauseit is closer to 50%. Doing so will yield the most conservative)JI1lplesize estimate and help guarantee that the sample error variance ofrourtes(\,llll beno greater than what you set as the maximum desired error.Forexample,the senior product manager at ACME Direct is prepared to

restanewsource of names from a list broker to determine the list's viabilityli3 sourceof new customers. On the basis of past experience, the seniorproductmanager believes the response rate for this list will be no greaterthan 4.75%.Ifshe requires the test response rate to be within 0.25% of theacrualresponse rate to expect in rollout with 900/0 confidence, how manyromes shouldshe test?

Inrhisexample,

p = 0.0475 (the estimated response rate to be achieved)

E = 0.0025

z = 1.645 for a 90% confidence level

Pluggingall known data into the formula provided yields the followinglample size estimate:

[(z')(p)(l - pi]n= E'

[(1.6452)(0.0475)(1 - 0.0475)[n=

0.00252

[(2.706025)(0.0475)(0.9525)]11=

0.00000620.12243060.0000062

1Z = 19,747

n=

Therefore,if the senior product manager tests 19,747 names, she will begilar d . ill fall wjthin3ntee 1 With 90% confidence that the test response rate w -!01i" f I ' . II t Once the test, • 100 the true response rate she can expect III ro au . .'fin I hi id b d . es ro assess this. a l t IS samplesize will give her the confi ence s e estrImas . hiIfaVia e source of new names. . I I f hertheseni d . h confidence eve 0enor pro ucr manager decides to Increase t e )

~mpJ 11USt she test.'response rate (from 90% to 95%), how many names I

Fora 9)"" f10 COn idence level,

z = 1.96[(1.962)(0.0475)(1 - 0.0475)]

n=0.00252

_____ r.-

300 OPTlMAL DATABASE MARKETING

[13.8416) 10.0475)(0.9525 )J0.0000062

0.17380830.0000062

n ~ 28,034

n=

To have more confidence in the results of her list test, all elsebeingequal, she needs to sample 28,034 names. Obviously, this is a majorincrease in the number of names to test-a 420/0 increase, to be exact.Ingeneral, the higher the confidence needed, the more names are required tosample.

If the senior product manager decides to keep the confidencelevel"90% but reduce the allowable error in the estimate from 0.25% to 0.20%)what is the impact of her decision on the sample size?

[11.645")(0.0475)(1 - 0.0475)].0022

[(2.706025) (0.0475 )(0.9525)]0.000004

0.12243060.000004

n = 30,608

n=

n=

n=

To have less error variance in the results of her list test, all else beingequal, the senior product manager will need to sample 30,608 names,again, a large increase in the number of names to test. The less variancedesired in a test estimate, the more names are required to sample.So, how do we decide what is the appropriate level of confidenceand

tolerable error variance (E) when planning rest sample sizes? There isclearly a trade off. The smaller the amount of error you can tolerate in yourestimate or the higher the level of confidence you need, the largerthereqUired sample size. It depends on the importance of the test and howaccurate the test result needs to be.

Determining the tolerable error variance. To determine the tolerableerror variance, the SenOr product manager must take into considerationhow close the expected response rate for the new list is to the "renrino-r.ent~' response rate level. The closer the expected response rate of thenewlist 1S to this cutoff level, the more important it will be to have a (estestimate with little error variance.

Assume the senior product manager in our example will promote anrnew prospect list as long as it meets her break even response raceof4.50%. If rhe senior product manager truly expects the list rest to yieldat

.....-------------------- ...rhoningand Designing Marketing Tests 301

leasra4.75% response rate, then for decision-making purposes she onlynteds(0 test enough names to ensure that her resuJting confidence inter-vaiisnowider than ± 0.25% (4.75% minus 4.50%). There is no need to"tire a narrower confidence interval results. All that is needed foramlionmaking purposes (rent the names or not rent the names) is to beertain thatthe response rate of the new list in roll-out will not fall belowmebreak-evenmark. Testing more names to yield less error variance andhencea narrower confidence interval is not necessary. She needs only to""nough names to yield the level of precision required to make her~ocision.Determiningthe appropriate confidence level. To determine the

appropriateconfidence level, the senior product manager will goihroughthesame steps outlined in Chapter 13 for determining the appro-mateconfidencelevel for confidence intervals and hypothesis tests. Foroample,if we assume the new prospect list costs $5.00 less per 1,000names rentedthan most other prospect lists, a 90% confidence level would~l'emappropriate. It is a less risky proposition to make an error in thismariothan if the list costs the same or even more than others.

~ampleSizeDetermination for theDifferenceBetween Two Sample Proportions

Whenyou want to measure the difference in response rates betweentwotestpanels in order to be able to accurately determine jf one testh~beatenanother test, employ the formula provided in this section todeterminethe required sample sizes. This is one of the most commonlyuledformulaswhen designing format, creative concept, pricing, and offertestslVhileneeding to compare the results against the control offer ".

Tocalculatethe sample size required when you are concerned. with thereliabilityof the observed difference between two sample proportlOnS, you"~dthefollowinginformation:

• TheProportion for one of the samples (P,) .. dAnestimateof the response rate of one of the test panels IS requITe

l·

T:b· . H you need a [air yIS can be based on prior i1'lformatzon. owever, . II

goodestimatefor one of the test panels to use the formula. Typica Yi.' d against a contraanewpackageor telemarketing script IS compare I

P k te 'or the contraacageor script. In all likelihood, the response ra I'

lIIill beknown, based on historical ;"formation.I Theminimum difference to derect as significant (d = PI - P2)

Yo" also need the minimum difference (d) between the twtO testr p . d with nertam y. Fores onserates (p1 and p 2) you requlTe to rea

�~ OPTIMAL DATABASE MARKETING

example, a control package is expected to yield a 4.00% responserate. The addition of a new premium is being tested, and at leaslfive additional orders per 1,000 nantes promoted are needed 10break even (or a 4.50% respollse rate). Therefore, the minimum di(-[erence of interest is 0.50%. In other words, the sample sizes forthese two panels must be large enollgh to ensure that a 0.50% dIf-[erence ill response rates is considered a significant difference(i.e., not 0).

• The proportion fat the other sample (pz)This will be calculated as p I plus (or mill us) the minimum differenceto detect (d) .

• The desired error rate a% and confidence level (I - a)%For example, a 95% confidence level allows ycu to state that tbeobserved difference between two test results of size d or greaterisstatistically meaningful with 95% probability. We recommend thatallsample size calculations be performed at a 90% or better collfidencelevel. Employing lower levels of confidence drastically increasesIberisk (error rate a) of obtainillg misleading results, which could leadtoincorrect and costly decisions.

Once all the information is known, the required sample sizes for thetWO

rest panels are calculated as

[zZ][(p,Hl-PI) + (pzHI-Pz)]dz

WherePland Pz are the estimated sample proportions for the two test panels

d ~ (PI - pz) represents the minimum difference to be detectedz is equal to 1.645,1.96, and 2.575 for 90%, 95%, and 99% confidencelevels, respectively (z values associated with other confidence levelscanbe found in Exhibit 13.1).

Note that, as previously mentioned to use this fotmula reliably yOUI ' I be]lllJmust rave a good estimate for at least one of the twa test pane s '

planned. Typically, when employing this formula, you are preparing [0 te51a new for' k If thiS is tbemat or creative concept against your control pac age.case, in all likelihood you will already know the expected response rate forthe control K . - d -f - I anel is nO

I. eep 1/1 nun ,I the response rate for this contra P f II

close to wh t bl ueressU ya was actually achieved you may not be a e to s ,detect the diff d " I as being" 1 erence 10 response rates between the rwo pane sstatistically significant (i.e., not 0). 'To furth -II D' t is teStinger I ustrare, the marketing director of ACME .rec eth ddi I ackag,e a men of a new scratch-off card to the current contro P 'I< (01

Suppose the control format is known to yield an order tare of 3.40 o

[(z)'][(pd(1-p,) + (/J,)(1-p,)Jd'

[(1.96)'][(0.034)(1-0.034) + (0.038)(1-0.038)](0.004)'

[(3.8416)][(0.034)10.966) + (0.038)(0.9621]0.000016

[(3.8416)][(0.032844) + (0.036556)]0.000016

13.8416)(0.0694)0.000016

0.2666070.000016

Planning and Designing Marketing Tests_~ __ ~-=--_----.::--~------ -...:'.303

34 orders per 1,000 names promored). If four additional orders per 1,000names promoted (or a 3.80% response rate) are required to cover theadditional costs of the scratch-off card, what sample sizes should thecontrol and rest panels be ro ensure that the marketing director will be ableto read ar least a 0.40% (3.80%-3.40%) difference in response rates as sig-nificant and meaningful with 95% confidence?In this example,

P, = 0.034 (the estimated response rate to be achieved)

d = 0.004PI = 0.038 (the value of P, + d)z = 1.96 for a 95% confidence level.

Plugging all known dara inro rhe formula provided yields rhe following

sample size estimate:

nt = nz =

", = n, = 16,663

n[ = n2 =

Tn order to conclude that 3.40% and 3.80% response rates aresignificantly different with 95% confidence, the markering direcror will be

required to test 16,663 names per panel.What will happen to the required sample sizes per panel if the marketing

director wishes to detect all even smaller difference as significant? The

required sample sizes would increase.So, how do we decide what is the appropriate level of confidence and

the critical difference (d) in response necessary to detect with certaintywhen planning test sample sizes? The smaller the difference to detect assignificant or the higher the level of confidence, the larger the required

sample sizes.Determining the tolerable error variance. To determine this, the market-

ing director must ask himself the following question: What is the minimum

304 OPTIMAL DATABASEMARKETING

difference in response that must be detected in order to make a reliablemarketing decision? In this example, the difference of Concern is 0040%,There is no need to ensure that a smaller difference can be read with statis-tical significance, He only needs to test enough names to yield the levelofprecision required to make a decision. Doing otherwise would be wastingtesting dollars.Determining the appropriate confidence level. To determine this, the

marketing director will apply the same steps outlined in Chapter 13 fordetermining the appropriate confidence level for confidence intervals andhypothesis tests, In our example, the new test with the addition of thescratch-off card is more expensive. As such, a 99% or 95% confidence levelseems appropriate. Because 99% is typically used only for the most extremeand risky tests, a 95% confidence level seems appropriate,

Marketing Test Planning SoftwareWith the use of The Plan-alyzer©, as mentioned in Chapter 13, you canalso determine appropriate sample sizes for given variances and confidencelevels, It will assist you in determining the appropriate sample sizeswhenyou are interested in the accuracy of a single test result or being able toaccurately measure rhe difference between two test results,

Alternative Testing Approachesfor Small Direct Marketers

Most major direct marketers test between 10000 and 20,000 namesperpanel for format, copy, and pricing test, Smaller direct marketers typicallytest 5,000 or fewer names per panel. As you can probably guess, on rhebasis of the information in Chapter 13, testing 5,000 or fewer namesper

I 'II ' f Viewpane WI not yield meaningful results from a markering point a 'Ho if h 1 folloWwever, J you ave a small test budget there are steps you calto hi' ' d our reste I' mcrease the odds of making correct decisions base on yresults,If, h~, you are a small direct marketer and simply cannot test enoug f I_

to ld si if , 'd r the aY.le sigru icanr results from a marketing point of View, consl e _lowmg id li h ibl For examgUI e tnes to ell' you make the best decisions pOSSI e, ,'fPie yo ' 0 h basIsa ,) u are testing a new and more expensive format. n t ecost I' h w formar,. ana yS1S, you determine that in order to break even on t e ne doverIt must generate an additional 2.5 orders per 1,000 names promote h'the control Th ' read sue, e required sample sizes that will allow you todifference in ' " h do YOU dolresponse as significant are too great. So w at

305Planningand Designing Marketing Tests

• Evaluate the test versus the control at the 90% significance level viahypothesis testing. If the test is 110t significantly different from the con-trol at 90%, evaluate the 10wet and uppet bound of a 90% confidenceinterval around the difference in test response rates. Then examine theupside and downside potential in terms of the difference in responsefor this test, taking into account cost and revenue figures. If the upsidepotential appears promising compared to the downside potential,retest to a larger sample. If the downside potential looks too greatcompared to the upside potential, drop the test and consider it a loser.

• If the test is significantly different from the control at 90%, based alla hypothesis test, check to see if it is also significant at 99%. If it is,you definitely have a winner and are advised to roll out with the win-ning tesr (assuming that the result of the test meets your marketingrequirements).

• If the test is significant at 90% but not at 99%, based on hypothesistests, retest the panel to a larger sample unless,• On the basis of an assessment of the lower and upper bounds of thedifference in response rates, you determine the risk of switching to beminimal, compared to the upside potential and, for example, you donot have any other options available for a new format. If this is thecase, then roU out and skip retesting. However, we advise you to con-duct a back test to ensure that the decision made was a correct one.

• On the basis of an assessment of the lower and upper bounds of the90% confidence interval for the difference in response rates, younotice that the test does not meet your minimum required differ-ence from a marketing point of view. That is, the difference inresponse based on the lower bound does not even come c1os~ t.omeeting your minimum required to break even on the test. If this IS

the case, don't bother retesting. Use the money saved here to test adifferent package with better potential.

When retesting is required, based on the above guidelines, .test eno,ughnames to read a significant difference from a marketing POint of vrew,Foll I . inanon auid Ii dined in this chapter, Atow t ie sample size determination gut e mes authO f Y have determined the testIS stage, you should not test roo ew names, aut b II iple and you ate nowo e potentially promising, based on a srna san ,ready to confirm based on a large sample. IF' f ersus the controOr example ACME Direct tested four new .ormats v

f' f 5 000 elected at randomormat. Each test was based on samples 0, names s Ifr h . Th Its of each test pane areom t e primary customer segment. e resushOwn in Exhibit 14.7. . % and 99%First, ACME Direct will conduct hypotheSIS tests at the 90 I diff t

Confidence levels to determine which new format is significaIExh

"y'bl 1e~e~f f h e shown III I It ..romthe control format. The results 0 t ese tests ar

306 OPTIMAL DATABASE MARKETiNG

Exhibit 14.7 ACMEDirect Format Test Series Results

Test Panel Sample Size Response Rate

Control format 5,000 100%New format I 5,000 175%New format 2 5.000 3.45%New format 3 5,000 3.29%New format 4 5,000 4.24%

Exhibit 14.8 ACMEDirect Format Test Series Significance Results

ACME Direct will then closely examine each format, based on theguidelines given to determine the appropriate actions to be taken.

New Format 1, This format is significantly different from the control a'.the90% level of confidence but not at the 99% level. On the basis ofthegwde·lines given, ACME Direct will assess the lower and upper bounds of the90% and 99% confidence intervals to determine the true upside and down·id . I ifid interval51 e potentia of switching to this new format. A 90% COIUI encereveals that the difference can lie anywhere between 0.16% and 1.34%.A990' f I' ywhere10 con idence interval reveals that the difference can re anb 0 18 . . ensivethanerween -. % and 1.68%. If this new format ISno more expthe control format, and ACME Direcr desires a new format for simplyah f . . h rhisnewc ange 0 pace, they should feel comfortable in rolling out Wit .f . . .... ( -018% Inannat. The maximum downside potential IS minimal . th. ) d . nsel if eyresponse compare to the upside potential (+1.68% 111 respo .. he'do roll out with this new format, a back test is advised. Orherwlse, I )should retest to a larger sample.

N F' . . I d'fferenl fromew ormat 2, This format was not found to be significant Y I. n'h I ME~a~t e comro at the 90% or 99% level of confidence. Had AC . dd . his Ollt( anucted this test to a larger sample size they could stop at t P ke'consid hi , d 'sed rotanSI er t IS test a loser. Because they did not, they are a VJ 'delia"closer look at the response rate received. On the basis of the g; 90%given, ACME Direct will assess the lower and upper bounds a aenli,1confide ice . I deterrni . d d wnslde potI rnrerva to etermme the true upside an 0

Significantly Different?

Test Panel Sample Size Response Rate (%) @ 90% @99%

Control format 5,000 3.00New format I 5,000 3.75 Yes NoNew format 2 5,000 3A5 No NoNew format 3 5,000 3.29 No NoNew format 4 5,000 4.24 Yes Yes

Planningand Designing Marketing Tests 307

of switching to this new format. A 90% confidence interval reveals that thedifference can lie anywhere between -0.16% and 1.03%. On the basis ofthis examination, they should retest to a Iaeger sample. It does showpotential, but nor enough to warrant a rollout like that of New Format 1.However, if this new format is more expensive than the conrrol format,

and, for example, an additional three orders per 1,000 names promoted ora 0.3% increase in response would be required to break even, they shouldnot retest. The bounds of the very liberal 90% confidence interval for thedifference suggests high odds of never meeting this requirement.

New Format 3. This format was not found to be significantly different fromthe control at the 90% or 99% level of confidence. On the basis of theguidelines given, ACME Direct will assess the lower and upper bounds ofa 90% confidence interval to determine the true upside and downsidepotential of switching to this new format. A 90% confidence intervalreveals thar the difference can lie anywhere between -0.28% and 0.86%.When comparing the upside potential to the downside potential, they caneasily determine that this test is not worth retesting and it should beconsidered a loser.

New Format 4. This format is significantly different from the control atboth the 90% and 99% levels of confidence. On the basis of the guidelinesgiven,ACME Direct will assess the lower and upper bound of the 90% and99% confidence intervals to determine the true upside and downside poten-tial of switching to this new format. A 900/0 confidence interval reveals thatthe difference can lie anywhere between 0.63% and 1.85%. A 99% confi-dence interval reveals the difference can lie anywhere between 0.28% and2.20%. 1£ this new format is no more expensive than the control format,.. IACME Direct is advised to rollout. With 99% confidence, it is guaranteecto beat the control by at least 0.28% in response. .If the new format is more expensive than the control and reqUIres some-

thing less than +0.28% in response to breakeven, ACME Direct should rollout. H the new format is more expensive than the control and reqUlreSsomething significantly more than +0.28% in response to hreakeven, they

should retest to a larger sample.

Regarding your most expensively created test panels-su~h as ne,:formats that may cosr you up to $10,000 or more to have deSigned-yosh ld I h firsr ti d This is true whetherau a ways test enough names t e irsr time aroun . - .Y

. k B d . you will be lesseningou are a small or large direct mar eter. Y omg so,th . k . D ' bl ding such test results. Ite [IS of misreading the results. on t ow rea I . .will be oui ." k p red to nllsreadmg orher

I e quite an expensive mistake to rna e com a .' . retest panels. The bottom line is this: If yOll are spending slgl1lfJcantly 1110

f ' ared to other testInoney On the creative development of a new annat con p hpanels, test it right and do not risk misreading the results. Test enoug


names to truly gauge the response correctly from the start. Not doing socould cause you to make a costly mistake.

Chapter SummaryTesting is the foundation upon which direct marketers build and grow theirbusiness. There are many important test design rules that must be consid-ered before implementing any test series. In this chapter, we discuss fivesuch rules. In addition, we discuss test design considerations regardingoutside list testing, including full factorial test designs. We also discuss theimportance of proper sample size determination. Without proper samplesizes, your test results will have error variance so great that rhe results willbe unreliable. We reveal several formulas for determining the appropriatesample sizes. Finally, we discuss alternate testing approaches for smallerdirect marketers with a limited budget.

Review Questions1. Discuss the five rules that must be followed to ensure that market-

ing test results will be readable, reliable, and projecrable.

2. Explain the advantages and disadvantages of full factorial testdesign.

3. Why is it important to determine appropriate sample sizes prior toany test execution?

Marketi ng Databasesand the Internet

Ever since Keri first logged on to a bookstore Web site and was given sug-gested readings, she has been intrigued with the possibilities of Internetcommerce. Over the years, she has registered on a number of sites and hasreceived targeted e-mail about a number of products. Like many Webusers, Kerf occasionally receives unsolicited, nontarget e-mail that she findsannoying. Kerf finds e-mail messages from direct marketing news servicesand Internet service venders particularly valuable for her work. These mes-sages are oriented to providing customers with information and developingrelationships rather than focusing on a hard sell. Kerl hopes that any e-mailPrograms she develops will provide value to her customers.

Ken wants to understand the advantages and limitations of Internet market-ing and determine which strategies and tactics would be most beneficial inhelping her reach her objectives. immediately; she recognized a potential costsavings. Not only was e-mail less expensive than paper mail, but she also sawpotential savings in order processing and customer service. She thought ofvarious forms of e-mail aymmunications such as newsletters, personalizedmessages, and messages with links and attachments. To develop e-mail cam-paigns, she would have to link the house file to the Internet. furthernlOre, todevelop e-mail campaigns for non customers, she would have to explore thesources of e-mail lists relevant to her p1'Oduct category.

Sophisticated data analysis and Web data mining would have to wait untilshe and Inside Source's analysts understood more about the systems andtools available to them. Keri's immediate objectiue was to establish anongoing contact with existing customers through e-mail and test some

alternative communication strategies.

In earlier chapters of this book, we presented information about Inrernerdatabase applications whenever appropnate. in the next rwo chapters,

309

310

we focus on Internet database techniques and applications. This chaptercovers basic database concepts as applied to the Internet, and Chapter 16delves into data analysis and mining.Developing a strategy for Internet database marketing is fundamentally

similar to developing strategies for other direct marketing media. Focusingon customer relationships and repeat business is more profitable over thelong term versus focusing on single transactions. Cross-selling is also animportant strategic option that applies to all direct marketing channels.Internet marketers such as Amazon.corn are attempting to implement thesestrategies. Amazon.com is building a large database that uses customerinformation to develop e-mail promotions that remind customers aboutproducts in categories in which they have demonstrated an interest.Furthermore, Amazon.com uses the database to cross-sell products rangingfrom books, CDs, electronics, health and beauty aids, and patio furniture.For example, book customers receive package inserts with discount codesfor purchases at a drugstore Web site. Cross-selling tactics also includeusing electronic coupons sent in opt-in (opt-in is a policy requiring individ-uals to give permission before an organization sends offers to them) e-mail.Therefore, collecting and analyzing data effectively and efficienrly is criticalto evaluating Internet marketing programs as it is with other directmarketing channels.

Database Integration

As Internet marketing evolves and marketers understand its advantagesand limitations, we anticipate that the Internet and other forms of elec-tronic communication will be considered a component of an integratedapproach to database marketing rather tban a separate enriry. Indeed,some companies are already taking an integrated approach. Customers areable to communicate and receive communications from a number ofsources., An integrated database allows multiple communications andtransactions to be tracked, regardless of whether by phone, mail, !nterner,wireless devices (m-commerce or mobile commerce) or at a physicalretail location. Indeed, rhe integrated clicks and mortar (aka clicks andbricks) approach is more common than single-channel markering on rheInternet. A study by the BOston Consulting Group of 400 online retatlers

showed that the clicks and mortar retailers accounted for 62% of onlinerevenues whereas 1"' f 38% The

, < pure on me retailers accounted or o •

multichannel retailers have an advantage over the single_channellnrernermarketers. They spend less to acquire customers have lower marketingexpenses, and have more repeat business (Book: 2000). Tbe successfultrend for these multichannel marketers held for the 2000 holiday season(DavIS, 2001).

Marketing Database and the Internet 311

A critical component in developing a responsive, customer-focused,multichannel marketing system is integrating or synchronizing databases.Although it is not the domain of this book to discuss the technical aspectsof database integration in any detail, it is worthwhile for the databasemarketer to understand the general concepts behind the integration and itsimplications fat developing marketing strategies and programs, Exhibit 15.1is a simple schematic diagram of the possible inputs into the database for amultichannel marketer.As we discuss the analysis of online data in the next chapter, you will see

that the ability of an organization to access data from multiple sourcesprovides a competitive advantage.An organization that does not integrate databases (e.g., using a separate

database fat mail, the Internet, telemarketing) loses efficiencies and valuableinformation about customers. This might result in marketing programs thatare not only inefficient but also may break down customer relationships.For example, calling or mailing an offer for a product that a customer hasalready purchased on the Internet may undermine the relationship with thecustomer. ("Why ate you bothering me with this' I already purchased one.")

Integrated

Telephone Database

System StrategyInternet and

Program

WirelessDevelop-

Customer ment

E-MailRecord

Updating

EXhibit 15.1 Possible Inputs Into the Database for a Multichannel Marketer


There is speculation that the higher level of abandoned shopping carts(shopping carts thar are left when the customer leaves the Web site aftermaking a selection bur before finalizing the transaction) for multichannelmarketers results from problems associated with integrating legacy system(older mainframe systems) databases with e-cornrnerce applications.Efficiencies can be enhanced further if supplier databases are also integrated.With this level of integration, customer order levels can automatically triggerinventory replenishing systems.One solution for database integration is to convert from the older legacy

systems to newer systems with a centralized database that can accept real-time inputs from a number of sources. This is nor a feasible alternative formany organizations that have large databases with millions of recordsstored on the legacy systems. In addition, legacy systems still work well, andfor many organizations other alternatives offer no advantages. However,most legacy systems were developed for batch processing rather than real-time processing (i.e., there is not an immediate change in the database as anorder is processed or a promotion is sent; the changes are made at periodicintervals). Technology companies such as IBM, Siebel, Kana, and Cayenrahave developed software and systems to integrate databases, making themresponsive to changing consumer channel preferences and organizationalobjectives for analysis and marketing program development.An example of an integrated multichannel marketer is Payless ShoeSaurce

Inc., one of the largest footwear retailers in the United States. Payless esrab-lished a Web site to offer customers more options for purchasing shoes. TheWeb site transactions are integrated with store transactions. Customers mayreturn or exchange online purchases to anyone of more than 4,300 PaylessShoeSource stores in all 50 states. Payless even allows items that are notcarried in a store to be returned there, and they also offer free shipping andhandling for online purchases when products are picked up at any stores("Customer Service," 2001). Other companies such as the GAP, CostCO,RiteAid, Williams-Sonoma, Wikes Lumber, and Ethan Allen have also movedto integrate online and offline databases to increase the effectiveness ofmarketing programs and enhance customer relationships.Despite the movement toward integration, the Internet has unique

characteristics and potential and deserves specific attention. We addresSthese distinctions in the following sections and consider how databases areused to implement marketing strategies and programs on the Internet.

Growth in Internet Commerce

Inter.ne~ commerce is experiencing a dynamic growth at this time. Growthpredictions vary widely. eMarketer (2001) summarized data from varioussources for both consumer and business markets. For 2002, in the b-to-b

MarketingDatabase and the Internet 313

market, estimates vary from $543 billion (Ovum) ro $6.8 trillion(Computer Economics). eMarketer predicted $841 billion. For 2003, theestimates range from $858 billion ro $9.9 triJlion. eMarkerer predicted $1.5trillion for 2003.In the consumer market, the estimates are also quite variable. For 2002,

they range from $81 billion (Ovum) ro $870 billion (Goldman Sachs);eMarkerer predicted $167 billion. The predictions for 2003 range from $133billion (Ovum) to almost $1.4 trillion (Goldman Sachs), and eMarketer'sprediction was $250 billion. Although there is variability in e-comrnercerevenue estimates, all analysts predict consistent growth over the nextseveral years.

We should emphasize rhar Internet commerce is still a small part of rheoverall economy. For example, in 2000, revenues of one large retailer likeWal*Mart exceeded the roral of all online consumer sales. However, manybusinesses are changing the way they conduct business because of theInternet. There are still numerous questions and challenges associated withthe new Internet economy. Some of the business models are not solidified,and many of the current Internet businesses will disappear or consolidate asthe industry evolves. Some initial approaches to developing an Internetbusiness have been questioned. For example, an approach used by severalInternet start-ups is to build large customer databases without an immedi-ate concern for profitability. They hope that the company will carry its largeCustomer database into other business areas. This strategy is dependent onCustomer loyalty and the capability of the company to cross-sell products rocurrent customers. Therefore, an essential component for the success of thismodel is the ability of the company ro use the database to develop loyalrelationships with customers. It is uncertain whether strong customer rela-tionships can be developed for commodity-type products that can be pur-chased from a number of sources. If the product is the same, only price andservice can differentiate products from the customer's perspective. Indeed,if low price is the most important differentiating factor for consumers, thenit will be difficult to establish brand loyalty. Consumers can easily movefrom one Internet retailer to another and purchase on the basis of price.Service however does seem to be an important aspect of Internet

commerc: for COnSu.~lers.More women are now online than men, althoughmen spend more time online (Pastore, 2001). More important, according tothe Ernst & Young 2001 Global Online Retailing Report, online shoppersin the United States are 60% female and 40% male. As the dnvmg force foronline growth, women will be demanding the service they have co~e toexpect from other direct marketing media and brick and morta~ retatle.rs.Therefore, online companies are making customer interactions With service

" A hi I centage of Internetrepresentatives immediately responsive- 19l per .I "C e t1y successful companiess 10ppers abandon their shopplllg carts. onsequ n , fII' "I' to questions or concerns 0se ..mg online need to be immediate Y responSIve

III

314

potential buyers on their sire. Research conducted by Harris Interactiveindicated that customers who contact customer service spend more thanthree rimes as much on average as those who do not contact customerservice. Customer service is particularly valuable to new online shoppers.Buyers who made their first online purchases in the previous year madeabout half of the customer service contacts ('tLivePerson and HarrisInteractive," 2000).

Responding hours larer bye-mail is no longer considered good customerservice on the Internet. Instant messaging through pop-up dialog boxes orvoice technology are ways companies are attempting to become moreresponsive to consumers. A number of companies such as LivePerson,eGain Live, and Cisco provide live customer service on the lnternet. Thesecompanies provide services to companies with e-business sites such asQVC, Lands' End, Neiman Marcus, Godiva Chocolate, and Block Drug.

Online customer service usually works in the following way. Customersneeding assistance click on a text hyperlink or icon. A window pops up,alerting a customer service rep. Customers are then prompted to input theirname so that the customer service rep can refer to them by name. The chatsession begins and customers ask questions and receive responses in realtime. The customer can still keep the Web site window open and to refer to

particular products or terms of an offer.The customer service rep can respond to the customer's questions

through the use of either common preformarted responses or a customizedresponse. This method allows one customer service rep to handle severalcustomers at the same time. Performance indicators (e.g., call length andinterval between question and answer) are available. Some of the live cus-tomer service software allows customer service reps access to a customer'shistory that can facilitate service in some situations.

The Internet Versus Other Database Marketing MediaBusinesses have already recognized several potential advantages to con-ducting business on the Internet. Cost savings are an important advantagefor Internet marketers. Online catalogs are much less expensive than papercatalogs that are delivered to customers. The cost of e-mail is also muchless than delivered paper mail. Online transactions eliminate the cost ofhiring order takers and reduce errors if orders have to be transcribed frompaper forms. E-mail costs have been estimated to be as low as $0.05 percontact, depending on Content development and list costs. Direct mail cancost several dollars per contact, and telemarketing contacts can exceed$10.00.

Many more items can be cataloged on the Internet than is feasible widlpaper or other media. For example, QVC, the television shopping network,

Marketing Database and the Internet -------315

hasa Web site that allows customers to select and purchase products online.It would not be feasible for most customers to watch extended hours ofQVC programming to be exposed to all these products. The Internet is animportant supplement to QVC's marketing because customers can-examineproducts that they missed when the products were featured on television.The customer can also access a product that they saw in the past on tele-vision but only recenrly decided to purchase. In general, the Internet is moreinformation rich compared ro other direct marketing media. Marketers canprovide potential customers with additional information that would not bepractical for other media. For example, business customers requiring moreinformation could click on a white paper, case study, or technical diagrams.Speed is another advantage of Internet marketing. E-mail offers can be

sent to thousands of recipients in a matter of minutes. New catalogs andupdates of existing catalogs can be put online without the need for theextensive production and mailing processes that can take weeks or months.Testing offers can also be expedited over the Internet. Promotions associ-ated with Web catalogs or e-ntails can be tested in real time as customersclick through to a sire and place an order.As technology advances, more e-commerce sites will have multimedia

capabilities, including giving customers the opportunity to view three-dimensional representations of products. Saturn allows visitors to their Website to take a virtual tour of the inside and outside of models of their cars(Saturn Corporation, 2001). Virtual tours are also used to market real estate,vacation destinations, and college campuses. Visitors to music commercesites can hear samples of CDs they are interested in purchasing. Thesemultimedia options are not possible or very costly with other dlr~ctmarketing media and allow customers to have a more personal contact With

a product.Customer service has potential advantages on the Internet if it is

Conducted properly. E-mail is less expensive than mail or phone calls forP

idi " inf t" 1 Some InternetrOVI mg customers With order status In orrna 101 •

marketers allow customers to check their order status online. This custom~rservice advantage is significant for b-ro-b marketers. Not only does thissave the cost of a human contact, but it is also a more rapid method formany Customers. United Parcel Service, for example, allows cust01~ers"toinput tracking numbers to determine the status of their package delivenes.G k tI f their order by enter-areway also allows customers to chec ie status 0. C ee whether the order

ll1g an order code or phone number. ustomers can s "h b " I b hi ped An estunatedas een received is in production, or las een s P ., id d d dina n the srarus of theOr actual shipping date is also provi e, epen 100 0

order.C. d fit -ner applications, hasISCO Systems a marketer of har ware or n er ""' f "I"" . Web site. Customers

extensive customer service and support act ines on ItS I " Id f e rC<Tisterfor tee mica

Can download technical documents an so rware, 0


workshops, manage service contracts online, and order products online.Cisco cites the following benefits of online ordering:

• It improves your company's productivity by have customers orderproducts and service online anytime, day or night.

• It knows where a customer's order is in the manufacturing andshipment process and how soon ro expect delivery.

• It provides up-to-date product quotes ro your company's clients byaccessing the most current price lists.

• It stays current with your company's accounts payable and measureshow expenses are tracking ro your budgets.

• It reduces the time it takes from submitting an order to receiving theproduct

Cisco's objective is to build customer relationships by providing respon-sive service and support systems. Because the majority of Cisco's customersare technology savvy, rhe most effective and efficient method to maintainrelationships is through the Internet. In general, many businesses find theInternet an excellent tool for conducting transactions and maintaining rela-tionships with customers and suppliers. The dramatic growth in h-to-bInternet commerce is probably due to the cost efficiency and responsivenessof the Internet for both business customers and suppliers.

In contrast to the b-to-b market, consumers usually do not have the sametechnical skills or purchase requirements as businesses. Therefore, consumermarketers have to develop different methods for Internet commerce. Goodproduct graphics, easy navigation, and immediate customer service responseare becoming more prevalent on consumer commerce sites. When customersContact a service representative, the service representative will have an advan~tage if a database can be called up that contains detailed infotmation on thatCustomer. For example, if customers have a question about a new productthey've seen on a Web sire, Customer service representatives may be abl~torespond better to the Customers if they have access to the customers' prevJOu~purchases and other information. The service reps could use the customersprevious purchases as a reference point for the new product under quesaoa,comparing colors, sizes, dimensions, content and so on. Knowing where

I· , hasCustomers rve can be helpful for a number of types of products sueclothmg, garden supplies, and auromotive parts. As we discussed ptevlouslytn this chapter the fli " ' ay reduce, use 0 rve text or VOICeonline customer service mthe abandoning of shopping carrs, and automatic queries and FtequentlyAsked Questions databases could facilitate the customer service interactionby providing custo ne ' h ' k '

I rs Wit a qUIC er response and more opnons-Anothet potential advantage to Internet marketing is the initiation of the

COntact. With telemarketing and mail the marketer usually initiates theCOntact. But with 'I keri .' d h contacte-rnar mar enng, the customer has inmate t e

Marketing Database and the I.nternet 317

by registering on a site and often allowing the marketer to send additionalinformation. In many circumstances, the customer initiates the contact byvisiting a Web site through the use of search engines or ads from othermedia. Customers who visit or return to a Web site are good prospects,because they have moved their interest to the next level of examiningproducts offered on a Web site. In other direct marketing media, offers areusually self-contained and require less effort by consumers. In consumerbehavior terms, the customer is often more "involved" in the process ofpurchasing online, and higher levels of involvement are associated withincreased purchase probability and brand loyalty. The Internet can give thecustomer a greater degree of control over the purchase process relative toother direct marketing channels.

The Internet also provides the opportunity for real-time analysis of dara.The marketer could evaluate critical factors such as the time needed forcustomers to find a product, the number of abandoned shopping carts, thepages that customers take as they move through the Web sire to purchase,and sources-c-e.g., Web address (URL) or Internet service provider (ISP)-of the customer. Managers could be alerted immediately when certainproblems are detected (e.g., a high percentage of people clicking pasr a newproduct rather than placing it in a shopping cart).Testing can also be facilitated on the Internet. Several variations of offers

can be tested and the results evaluated quickly, allowing poor offers to beeliminated as soon as possible. With other direct marketing media, the devel-opment of alternative offers for testing (e.g., mailer modifications, catalogchanges) would be more costly and time consuming. We should emphasize atthis point that all the advantages of Internet marketing that we have presentedin this section are dependent on an efficient and effective database system.

Limitations of Internet Marketing

Although marketing on the Internet has distinct advantages, it also haspotential disadvantages. Some marketers are concerned abo~t the c?n~r~of offers and promotions on the Internet. Pictures and graphiCSare limiteby Customers' equipment and their Internet connection. I.n contrast, formail, catalogs and television the marketer usually has more control over

d . ' , I d" h olurion 01 pictures ofrepro ucnon, Because of down oa nrne, teres Iproducts on the Internet is lower relative to printed pictures in the .caraogindustry. In addition, customers might attribute long download nmes oraccess problems with a site to the marketer, regardless of the source of the

bl d beyond the control ofpro em. Sometimes the problems are ue to sourcesth " their ISP In the future,e marketer such as the customer s equipment or - .

db d h I ies (e g digiral, cable),as CUStomers move to faster broa an tee no ogr . -,these problems will be minimized.


Although many consumers find non targeted mail (junk mail) a nuisance,nontargered e-mail (spam) is often considered an invasion of privacy.Therefore, marketers must be extra cautious when developing e-mailcampaigns. As we mentioned previously, many marketers are adopting anopt-in only policy for e-mail lists to minimize the chances of alienating con-sumers. When marketers do make a mistake and e-mail offers to Customerswho do not wish to receive the information, the customers should be able toimmediately have their names removed from the marketer's list. In somecases, even marketers who are sensitive to customers' preferences have madeit difficult for customers to remove their names from a list. Sometimes e-mailrecipients have to click through several links and wait for several pages toload before they can request name removal from the e-mail list.Target appropriateness is another factor and potential disadvantage. Natali

potential customers for a product have access to the lnrernet, For example,marketers who target older consumers may still need traditional directmarketing media, because this target group has less access to the Internet.Furthermore, according to a study commissioned by the American Associationof Retired People (AARP), older U.S. citizens have serious concerns abourInternet commerce and don't feel they have appropriate computer skills. Thisis another problem with Internet marketing that will probably be resolved overtime. Technological advances will make Internet access easier and safer.Furthermore, today's Interner savvy younger and middle-aged consumers willbe the older consumers of the future.

Previously, we mentioned that in many circumstances, the customer ini-tiates the contact in Internet transactions whereas in other direct market-,ing media, the marketer initiates the contact by calling or mailing an offerto the prospect. There is also a downside to customer-initiated contact.Marketers must make their Web sites easy to find with search engines. Inaddition, many marketers don't want to wait and attempt to drive traffic totl.1esite with ads in other media or through partnerships with other Internetsites such as portals. Marketers are still attempting to work out the bestmethods to drive traffic to their Web sites. Recently, markerers have had toreevaluate the effectiveness of using banner ads and television commercials.According to NielsenlNetRatings May 2001 data click rates are a meager0.49% for banner ads. Furthermore, many co~panies have abandonedmass t~lev,ston advertising as a means of making customers aware of the~rWeb sites. At the core of the issue of driving tra£fic to the sire are basiCprel~lses of good marketing communications-that is, find alternanvemedia that are viewed and attended to by the target group. Does it makesens.e to lise banner ads or television commercials to drive traffic to a site Ifthe mtended target group does not click on banner ads or pay attention totelevIsion?

Which is the best way to drive traffic to a Web site? Citing data fromAndersen Consulting, Weaver (2000) concludes that banner ads are better

Marketing Database and the lnrernet

(25%) for driving customers to shop online compared to newspapers andmagazines (14%). Television had the lowest ability at 11%. Weaver claimsthat the effectiveness of the banner ads increases when they are targetedusing technologies such as Double Click or Engage.Consumers have some generally negative perceptions about shopping On

the Internet. The results of a study conducted by Drozdenko and Cronin(2001) indicated that consumers rated physical stores as having higherproduct quality and service quality for several product categories comparedto online stores. In addition} consumers in the study rated a lower purchaseprobabiliry and safety for online versus physical stores.

In a second study, Drozdenko and Cronin (2001) found that marketerscould improve the perceptions of Inrernet shopping by providing addirionalinformation to consumers. Reassuring consumers about the security ofpersonal information, safety of credit card transactions}easy return policies}andresponsive customer service nor only can improveconfidence in purchasesafety bur also improve the perception of the product's value, quality, andpurchase probability.Consumers, public interest groups, and legislators seem to be more

concerned about the privacy of data collected on the Internet than to datacollected through other direct marketing media. Privacy on the Internet is acomplex issue even for the U.S. government. There arc concerns thatgovernment Web sites are inadequate in protecting consumers. A number ofprivacy violations were cited in a report submitted to Congress by 51inspectors general. These included violations of establisbing privacy policiesfor specific government Web sites and violations of the Children's OnlinePrivacy Act (Bremner, 2001). We examine privacy issues in more detail inChapter 17.

Personalization: The Great Promise of the Internet

The great promise of the Internee is for one-to-one marketing and perso~al~izing relationships with customers. Customers can receive personalizedinformation on the Internet in several ways. E-mail can be tailored to thepersonal preferences of customers based on past transactions or other cus-tomer data such as demographics and psychographics. Web pages (portals)can be personalized with news and information that an individual prefers.For example, Yahoo! earns revenues in a number of ways, such as byaccept-ing advertising developing and hosting Web sites, and charging merchants. ' . I TI f it is important thata revenue share on sales driven by Ya 100! iere ore, I

Yahoo! entices people to come to their site often. Yahoo! allows people to

d I . h M Y I I Content and layout caneve op personalized Web pages wit Y a 100. .be Customized to personal tastes. Sports, business} entertaimnent, headlinenews, and health are some of the categories that can be selected. Updates on

319


certain stock prices, sports team scores) weather, and breaking news inspecific categories can be presented when the page is called up.

Cookies are another means to personalize a customer contact on a Website. Cookies are files placed on a person's computer disk by a Web site'ssoftware when that person visits the site. Cookies contain informationabout a repeat visitor to the site, thereby eliminating the need to reregisteror sign in on Web sites. When the browser connects to a site, a cookie isread from the disk of the visitor ro rhe site. If the cookie is not there, thevisitor has to reregister or sign in. Furthermore, any Web pages that thevisitor has personalized (such as My Yahoo! discussed previously) wouldno longer retain the custom settings.Cookies can also be used ro track browsing patterns of individuals. This

is currently a controversial area) and some public interest groups are con-cerned about organizations planting cookies and then retrieving them insuch a way that allows them to build detailed profiles of rhe interests,purchasing patterns, and lifestyles of individuals. It appears that howInternet data on individuals is collected and used will continue to be an areaof concern for years to come.

Technological advances in wireless communications include Internet-enabled cellular phones that provide the possibility to engage in mobilecommerce. According to the Strategis Group, mobile device use in theUnited States will grow from its current 2% to 60% by the year 2007.Jupiter Research estimates that by 2005, worldwide m-comrnerce revenueswill reach $22.2 billion. However, predictions about m-comrnerce arequestionable because of many unknown factors of the new medium,according to eMarketer analysts (Blank, 2001).The new wireless technologies promise to bring another level of person-

alization. Using a database containing an individual's shopping preferencesand global positioning satellite (GPS) systems, retailers have the potential todrive traffic to brick and mortar locations. GeePS.com, for example, cansend promotions, product information, price comparisons, and so on toowners of Internet-enabled cellular phones as they move into the proximityof a retailer (Dana, 2000). IBM anticipates that m-commerce will expandto the global level. Using an enabled cell phone, consumers will be able tomake purchases at locations in other countries without the need forcurrency or credit cards.

Privacy will also be a key concern of wireless Internet marketers. Ifcons,umers receive unwanted or excessive promotions from marketers, theyare likely to turn off the wireless device Ot complain ro the service provider.Some wi I d .. I 'f' non. ~e ess a vernsmg p acernenr companies are removing identi rca 1

inforrnario-, (names, e-mail, add tess, etc.) from personal pr-ojiles using onlya randomly generated number to identify the wireless device.

Although privacy, device limitations, and slow networks are barriers tom-commerce, consumer apathy may be the greatest threat to its growth,


according to Jupiter Communications. Fewer than 0.1 % of 110 millionU.S.wireless users purchased products using wireless data services in 2000,and just 7% of these users intend to try mobile shopping in the next year.

E-MailMarketing

According to an eMarketer study, expenditures on e-mail marketing in2001 are expected to increase 110% to $2.1 billion (Tomasula, 2001).Although the availabiliry of e-mail lists is growing, there are fewer e-maillist providers compared to mail and telemarketing list providers. Mail andtelemarketing list providers also have a wider range of response lists rela-tive to e-mail list providers. Furthermore, fewer names tend to be on e-maillists as compared to mail lists. Some traditional primed mail list companieshave 100 million households in their databases that can be segmented by anumber of demographic and psychographic variables.

In 2000, most e-mail list providers had less than 10 million names on file.However, the number of names and categories on these e-mail lists is expectedto grow rapiclly in rhe next few years. For example, as of 2001, XactMaiJ hasa database of about 50 million opt-in names, and PostMaster Direct. com has30 million opt-in names in more than 3,000 topical lists.

E-Mail Applications

A number of e-mail tactics can be used to strengthen relationships with cus-tomers or acquire new custol11ers. These include

• Informing customers of new products• Offering incentives (e.g., discounts, special offers) to customers. .• Contacting a targeted group of noncustorners who have some sundar

characteristics to current customers .• Providing periodic (daily, weekly, monthly, etc.l information to

f I b . fs or other relevantcustomers in the form 0 news etters, news ners,informa tion .. I h h .. ted your Web sire, trade

• Following up contactS to peop e w 0 ave VIS!show booth, and so on

• Updating customers on order status

.. . ·1 to applications for otherAs you can see e-mail applicatIOns are S11111 ar ..

d· ' . k h elevant to the recrpienrIrect marketing media. The goal IS to rna e t em r .and elicit interest purchase, and brand loyalty. . h dI

. . '. . b t-in lists and lists t at 0t IS Important to differentiate etween op.. fil .. t to receive messages rom

not have the permission of the e-mal reclplen

322

list users. Nonpermission e-mail lists can also be acquired. Some e-maillist providers use computer programs that harvest e-mail addresses fromWeb sites. The term spamming refers to e-mail campaigns that are sentto recipients who did not specifically give permission for commerciale-mail of a specific type to be sent to them. For example, people mayhave given permission (opted-in) fat their e-mail address to be sharedwith companies that sell products related to skiing. These people areinterested in receiving information in that product category. However,these same individuals may have registered on a nutritional Web site butdid not give permission for their e-mail address to be shared with mar.kerers selling nutritional products. If these people receive e-mail from amarketer selling nutritional supplemenrs, it is considered spam mail. TheOwner of the nutritional Web site may have shared the e-mail addressesof people registered on the site without their permission, or anothercompany may have used software to extract e-mail addresses from thesite. We discuss the ethics of e-mail marketing techniques further inChapter 17.

E-mail campaigns are believed to yield response rates that are higherthan the average 1 % to 2% response levels for mail campaigns. Responserates for e-mail campaigns have been reported to be between 5% and 10%(Priore, 2000). Higher e-mail response rates are commonly reported. Forexample, Kawasaki MOtors said 28% responded to their sweepstakese-rnails (Cruz, 2001).

Although these response rates for e-mail marketing are impressivecompared to traditional printed mail, they may nor represent an accuratecomparison. Many companies that e-mail are adopting an opt-in list policy,but most printed mail lists are Opt-out lists. Opt-in lists have a higher prob-ability of COntaining good prospects in a particular product category.Furthermore, '<response" to an e-mail campaign is sometimes defined as a

"click-through," whereas response to traditional mail or telemarketingoften refers to placing an order.

E-Mail Formats

A number of formats are used for e-mail marketing. At a basic level, thesame e-mail message can be sent to all recipients on a list. This is the cheap-est and quickest method. However, because of the lack of personalization,software filters set up by the recipient or the ISP may screen out this typeof message.

Personalized messages are also possible. Informat.ion from a database~llcb as first name, product ordered, inquire date, and so on can be insertedInto ~he message to make it more relevant to the recipient. Personalizede-mail messages are shown in Exhibits 15.2 and 15.3.

Marketing Database and the Internet

Exhibit 15.2 Personalized E-Mail That Uses Past Purchase Data

To: [email protected]:Subject: "A Traitor to Memory" by Elizabeth George

As someone who has purchased books by Elizabeth George, you might like to knowthat her newesr book, "A Traitor to Memory," will hit the shelves June 26, 2001.You can pre-order your copy by following the link below:http://www.zzzzzzzz.zzzlexedobidos/ ASIN/05 538012 79/ref=mk_pb_sbr

Classical music, cybersex, and vehicular homicide figure prominently in this sprawl-ing epic, the latest in the bestselling Thomas LynJey series that has won ElizabethGeorge an enviable following on both sides of the Atlantic. This can only add toher growing repuration as doyenne of English

mystery novelists. -From Publishers WeeklyTo learn more about" A Traitor to Memory, n please visit the following page at:http://www.ssssssssssss.sss/exec/obidos/ASIN/0553801279/ref=mk_pb_sbr

Happy reading,Editor, Mysrery & ThrillersPS: We hope you enjoyed receiving this message. However, if you'd rarher notreceive future e-mails of this sort frorn ssss.corn, please use the link below or clickthe Your Account button in the top right corner of any page. Under the YourAccount Settings heading, click rhe "Update your communication preferences" link.http://www.ssssss.sss/your-account

In addition, e-mail messages can contain attachments such as document,picture, or sound files, and links ro Web sites that the recipient can use toplace an order or obtain more information.E-mail marketing can also be in the form of individual offers, newslet-

ters, Or periodic interest pieces. Some trade publications such as DMNews have daily e-mail updates of news relevant to direct marketers. Thesee-mail messages contain advertisements from companies in the industrythat the recipient can click on for further information.Other companies offer daily e-ntails to consumers on a topic of i~ter-

est such as word of the day, recipes, trivia, or quotes. The objective I~t,Okeep the organization that is sending the e-mail accessible to the recipi-ent. The e-mail recipient may click on a link embedded in the messaged b .' . offers or banner adsan e sent to a Web site that In turn may contain ,I '1 ., F example a foreignre evant to the interests of the e-rnar recrprent. or "

language site offers an e-mail that contains a word and pbrase of the dar'In addition to the word and phrase, links are also included III rhe e-rnar .The links might take the e-mail recipient to a site sellmg language orculture CDs.

323


Exhibit 15.3 Using E-Mail to Confirm Order Shipment

To: [email protected]:Subject: Shipment Confirmation

Dear Ralph,

Thank you for shopping at ZZZZZZZ. Please keep this email invoice for yourrecords.On January 26 we shipped your order number 3333333335 for rhe following irem:Santana: SupernaturalFormat: CD Quantiry: 1 Price: 12.58 to the following address via U.S. PostalService:Ralph Wakely78 Main St.Any town, CL 07777

Total number of items: 1Subtotal: $12.58Shipping: $ 2.99Sales Tax: $ 0.00

Shipment Total: $15.57

Your order was billed to Ralph Wakely. This shipment completes your order andis paid in full.

Most orders arrive within 4-8 business days. However, in rare instances it maytake up to 2 weeks.30% OFF Music by rhe Artists on Your Custom CD!For complete information about your order (number 3333333335) or to confirmthe status, click or copy/paste this link into your Web browser:http://zzzzzzz.comimyorder/otid=3333333335You can also access your Order History directly from our home page.Piease do not reply to this email. If you have questions about your order rhar arenor addressed in your online Order History, please visit our Contact zzzzzzz pageusing this link: http://sssss.com/serviceThanks again for your order.Sincerely,Customer Service

Chapter Summary -------------The rapid growth of Internet commerce has been fueled by a number offactors, including cost savings, increased efficiencies, and greater custorn~rcontrol. As more traditional database marketers move online, there IS


increased pressure to develop integrated database systems that allowcustomers to access an organization from a number of media. Integrateddatabases have advantages for both customers and marketers because cus-tomers are able to obtain product information, order, pay, pick up, and returnmerchandise from a number of locations. Marketers increase their value totheir customers by providing them with these shopping options and at thesame time gain additional insights into their shopping patterns with theintegrated database. However, integrated database development can be com-plex, especially when the integration involves legacy systems. Personalizedcommunication is a key advantage of Internet marketing. Compared to otherdirect marketing media, Internet personalization can be dynamic, changing asCustomers access various types of information and make new purchases.Furthermore, personalization on the Internet is expanding into the develop-ment of real-time customer service systems that utilize the database.

Review Questions

1. Discuss the advantages and limitations of Internet marketing.

2. Explain the concept of integrating databases across direct marketingchannels.

3. What are the customer service possibilities on the Internet?

4. What methods are used to personalize communications on theInternet?

5. What are some of the customer variables that are collected on theInternet?

6. What are the differences between traditional print and e-mail lists?

Analyzingand Targeting

Online Customers

After attending a workshop on basic online marketing, Keri wants tounderstand more about the similarities and differences between online andoffline database marketing. She realizes that there are different types of datathat can be collected about online customers such as tracking visits to aWeb site. Keri is also interested in software and services that track customerbehavior online. Now she wants to take the next step and start toimplement some of these techniques on the Web site of Inside Source. Inparticular, she is interested in testing offers and driving her currentcustomers to the Web site.

Maximizing the use of a-commerce customer data employs the sameprinciples and techniques as for offline enterprises. We covered these

techniques in Chapters 6 through 14011 data mining and market testing.In this chapter, we present techniques and stl:ategies unique to onlinebusinesses and their customers.

In particular, in Chapter 15 we noted that online customers initiate theirContact by visiting a site and providing an e-mail address. This permission tocommunicate and the economies of communication afforded bye-mail havecreated a revolution in the direct marketing business. In this chapter, weinvestigate how Internet communication changes the marketing dynamicand how information can be leveraged to exploit this new dynamic. We dis-cuss the similarities and differences between online and offline customers,the nature of the data available from online and offline commwlicarions, andhow to effectively leverage the data to gain marketing efficiencies.

Data Collected via the Internet

JUSt as the Internet allows customers to initiate a dialogue with themarketer, the Internet also allows customers to state their preferences and


provide a profile of information. Many sites have registration areas inwhich visitors are invited to outline their interests and their demographicprofiles. All reputable companies also have their privacy policy clearlystated on their Web site wherever they collect customer information.Many lnternet customers understand that by providing informationregarding preferences and interests, they will receive offers tailored specif-ically for rhem. They further understand that if they decide in the futurethat they no longer wish to receive communications, they can express thatpreference at that rime.Individual-level customer dara that are collected from a Web site can be

classified as either registration, behavior, or source data.

• Name• Title• Business title (if the site is a b-to-b site)• E-mail address• Postal address (business or home)• Phone number• Fax number

• Age• Income level• Gender• Competitive product usage• Current consumption level• Product attributes

Registration Data

Many e-cornmerce sites require consumers to register before they can makepurchases, get information, or enter the Web site. The type of data collectedvaries from site to site. Examples of Internet registration data typicallycollected include

Some examples of e-cornmerce registration pages are shown in Exhibits16.1, 16.2, and 16.3.

Exhibit 16.1 shows the Clairol registration page (www.c1airol.com). Themost comprehensive of the three pages displayed, questions on rherrregrstranon page contain many product usage questions, including brandprefere~l~e, usage frequency, and loyalty questions. .Exhibit 16.2 is a registration page from Saab USA (www.saab.com).ThlS

page asks only for name, address, and e-mail information. To help thembetter understand their visitors' intent, Saab should ask about their drivinghabits, If they ever previously owned a Saab, what type of car they currently

Analyzingand Targeting Online Customers

free 0 erfor haircolor us:ers:

y~u),IIMItd 10 ''''flW , FREE albsMpIi/JII 10 CcIq, SIluro,.(.fa JlZme I,omClfIIOf.--yfw, b-.sf JfSqUIli. fo, IhI "'-.sf n~ lips MIl i<M/;e .c.ldhanr;olflf tnd ~,. Y~ull ,,,,,iwlll.fttISlItllOf/Jtilli""'1OI1f ~9Ic$,~hlll~m"" CJfII?' up.1ts, tb$D/IIt~;REE, plusspKJai PI.moti,,/wQ/f'lSlesvvtd udJsNtlyf~, yeu! To gil C~IOISIlIlll" 1I'dIJding, $1 00CWPM' Mill ytlilf filS 1SSlJ'. pl.ase c~""II/'f1>I/ .wbmt IIwsSlI/Vfy ..

Only '''''fqUI5f P" hotJSIhold. g.oup 01~,gtrliztlNm.Offer good onlyIn the U.S. 10 "'''SlIb::tib.lSo~'Ill•• g' gf 13 Yw TtlM111e 13 OI~'

:::"::.';:'f0Y:'u~lJfW:W::':::/'fSf !la~'Pfflnf 01gv~

1. bthlsyour/i,""visll",""llOr.oom1

ryo;rM

Exhibit 16.1 Clairol Registration Page

drive, how soon they are looking to purchase a new car, if they plan to buy;.Iease their next car, and what other makes of cars they are considering.d ith such information, Saab could direct the visitors to the appropriateealer or provide a more targeted offer. However, as most direct marketersthe well aware, there is a trade-off-the more questions asked, the lowert e response. If this is the reason Saab is not asking more quesrions of theregistrants, we hope they are at the very least gathering some valuablechckstream data based on where the visitors go within the Web site (a click-stream is the sequence of clicks or pages requested as a visitor explores aWeb . ). Site. Doing so, for example, Saab could capture the fact thatregiStrants click on the page that discusses the details of the new 2002 Saabconvertible models. In this case, Saab would have very valuable informa-tion to help them properly communicate with the new registrants.Exhibit 16.3 is the registration page for Club Med (www.clubmed.com).

Club Med, like Saab USA, only asks for the visitors' name and address.However, they do offer a newsletter. Again, if they cannot ask questions onthe registration page due to a decrease in response, they can still collectvaluable information via c1ickstream data. This would provide Club MedWith very valuable information about the different vacation packages the

329


*Seab InlOlmalum.Hequest a Brochure Netecepe ~ '

Saab r • <.I .... window

Request a Brochure

To requu! furth., Info.motlon, pin,. tompt.tl tho fo,m bllow.

Broclmrerequost•• Mlred b.fo,. Octobt. I.2000 wmb. mtlld I 2000 Mod,l VH.brochu, •. On Octobor I, 2000 "I will blgl .. moiling nno'"quuu the 200 I Mod,l Vu,b••du ...

Plus. enid ... , f'MS,rMrJ,rMr•

Fl'nNlml ~

"' r---L•• tNamo *1 r-----------;Addrus"- rl----------,

City'"

SUtt >I- IS.I"'t... :::JZip" r----i-r-P~n. I ~

tcrreec Ho:tjJTl~eJeJer'lCes.cOlll COOl.aeled, W.illrlllij ==:J~ o-a '2lfJ 9 ~

Exhibit 16,2 Saab Registration Page

registrants read about while they are on the Web site, With this informa-tion, Club Med would be a strong position to send targeted offers,

Behavior Data

In Chaprer 15, we discussed the use of cookies as a means of personalizinga Web site, Cookies can also be used to track the parts of a Web site rhatare visited. As we also discussed in Chapter 15, a cookie is stored on theWeb visitor's hard drive with an identifying number. The cookies and thenid ifvi b b 5 so thatI enu ymg num er are passed by Web servers to Internet rowser,

I' , , ki b Isiror caneae 1 time a VISitor returns to a server that delivered a coo re, r e vbe recognized.

Bid h id '£y' inforrnation srored i h cookieeSI es tel enti IJ1g number, other information store In t e .

' I des rirec. , , ddi t recordmgInc u es previous pages vrsited and expiration date. In a mon 0' , ki I d that aVISltS, coo res a so permit a record of behavior to be store so

Analyzing and Targeting Online Customers ]31

~"~'~.~"~_~'~'~,""~~!!~H~..~•••••••••••••••••••••• liiYl1';l~l~l~

'l '.

II a!>Ot/t you

Sign uplIf wou ""antrO reteive special offers and news by e..mail, lell ""who you ale.Plust filln lha fallowmgform

Your e..mall (\oIIIt1l...uall;(>be l"lUlloglll) IYallrl,_old: rl------Ple""l1 canOrm yu"r p_.d: I;~~:onweWlIllKkYOUlfYOUIOlISe'fllUI tl~~;:;;;;;;::~The eK"eded a~r: IIf you fosry1llll p_old,~YOllr member 10 (d yOu 1\8'<1 alread, eeen fo ClubM,.

IMISS ::::JLa$! Neme: r--~'----.Firstr-------' Neme:Phone: SlIalll:

Tlllll:

Exhibit 16.3 Club Med Registration Page

CUstomer's page views can be identified. This record can be aggregated by theunique identifier found in the cookie to establish patterns of behavior for eachonline customer. The clickstream data can be valuable in many respects: Adirect marketer can determine the most popular pages viewed by each visitor

,

segment customers for offers, or use it to predict a likely response to offers.When a registered Web visitor becomes a customer through a purchase at

an e-commerce site, additional behavior information is available in the formof product preference, delivery choices, frequency of access to account infor-mation, usage of online discounts, "wish lists," and other loyalty devices.

Types of customer behavior data that can be captured includes

• Visits• Total page views• Specific page views• Time spent at the site• Products purchased• Customer service requests• Access to personal account information

• Discounts used


Source Data

The importance of source information for pro pet evaluation of anyacquisition effort cannot be stressed enough. The marketer needs to knowwhat media convinced a customer to visit the site.In their "Holiday Shopping Exploration" report, Digitrends quantifies

the various influences that contribute to a consumer shopping online:

• 42 % of online buyers shop on the Internet as a result of a promo-tional e-mail

• 41% shop online as a result of an article or site link• 39% are guided by a search engine• 38% are influenced by word of mouth• 29% are directed by an online ad• 27% are familiar with a site's offline store• 21% shop online as a result of a magazine article• 19% are influenced by TV or radio commercials• 17% are moved by print ads• 12% are persuaded by direct mail

Most offline direct marketers routinely track the source of a customer toevaluate the effectiveness of various media or lists and to correlate memberquality by SOurce. The source of a customer is just as important on theInternet for the same reasons. The method of collecting source will involveone of the following:

• Construction of unique URLs for each link or online ad• Construction of unique URLs for specific e-mail lists or direct mail lists• Evaluation of log files to identify search engines referring customers.• Collection of survey data at registration for customers who weredriven by print or broadcast ads, magazine articles, or word of mouth.

Understanding Internet Users and Online BuyersBefore proceeding, we must make a distinction between Internet users andonline buyers. An Internet user is anyone using the Internet for purpose: ofconducting research or making purchases or both. An online buyer IS asubset of Internet users. So what does an Internet user look like, and how~any are there? An Internet user is younger, has a higher income andhigher education than the general population. In particular,

• The number of Internet users aged 18 or older has been estimated tobe approximately 93.2 million by the U.S. Census Bureau as of

@5

Analyzingand Targeting Online Customers 333

August 2000. Other research suggests a higher number of users in thisage group: Harris interactive sizes the 18 + market at 114.2 millionas of June 2000, Pew Research Center estimates 113.8 million as ofAugust 2000, and MRI estimates 112.9 million as of spring 2000.

• MRJ estimates the number of adults (aged 18 at older) active in thelast 30 days at 86.3 million as of their spring 2000 survey,

• Estimates of numbers of Internet users who access the Internet fromwork range from 26.8 million by Media Mettix to 34.8 million byNielsen/NetRatings in August 2000. Morgan Stanley Dean Witter'sestimate was 36 million at midyear 2000.

• According to Forrester Research in January 2000 and the U.S. CensusButeau in August 2000, Asian Americans account for 3.8% of thepopulation but comprise 6.1 % of the Internet population. HispanicAmericans represent 11.6% of the U.S. population and 12.7% of theInternet population. African Americans represent 12.2% of thegeneral population but only account for 9.4% of the Internet users.Caucasian Americans represent equal numbers in the generalpopulation and the online community at 71.1 %.

• According to Pew Research in August 2000, 57% of Internet users areunder the age of 34. Only 17% of individuals over 60 years of age

have access to the Internet.• According to Media Metrix in June 2000 and the U.S. Census Bureau,roughly 58% of Internet users have household incomes of $50,000 ormore. This is in contrast to 46% of the general population of U.s.

households with incomes of $50,000 or more.• The UCLA Internet Project reporrs from their spring 2000 survey thatInternet penetration is greatest among well-educated, high-incomehouseholds. Eighty-six percent of individuals with a college degree oran advanced degree have Internet access. Seventy percent of individu-als with some college have access to the Internet. Among individualswith no college education, the Internet penetration drops to 53%.

How do online buyers differ from Internet users and the general popula-tion? Online buyers are younger and more affluent individuals who havebeen active Internet users for many years. Their unique charactenst'P 111

contrast to Internet users and the general population are as foUows:

• According to the Gartner Group in midyear 2000, the percentage ofindividuals with an income level of $50,000 or more who have pur-chased on the Internet is 51.2%. Only 39% of individuals withincomes less than $50,000 have made purchases online.

• Also according to the Gartner Group in midyear 2000, 36% ofInternet users aged 55 and older have purchased something on theInternet. This contrasts to 49% of Internet users in the 35 to 54 age


group who have made a purchase online and 46% penetration ofonline purchasing in the 18 to 34 age group.

• The UCLA Internet Project found a correlarion between tenure ofInternet use and online purchasing penetration: 71 % of individualswho were Internet users for 3 to 4 years had made a purchase online,whereas online purchase penetration was 26% among individualswho had been using rhe Internet for less than a year.

• Esrimates quantifying the dollars spent online in 1999 vary widely.The Direct Marketing Association estimates that the average onlinepurchaser spent $559, the Boston Consulring Group estimated onlinepurchasers averaged $460, but Ernst & Young estimates the averagevalue of online purchases ar $1,205.

Web Site ReportingMore and more direct marketers are establishing their presence online.Surprisingly, not all gauge the effectiveness of their Web site. The DirectMarketing Association commissioned a study on the state of the interactivee-commerce marketing industry in January 2000. Key trends in Internetusage among direct marketing companies include the following:

• About 96% of respondents reported the use of the Internet formarketing and sales applications in 2000. Sixty-two percent reportedrhe use of e-mail marketing in 2000.

• About 97% of respondents reporred that their companies have a Website. In 1998, 87% of respondents reported rhat their company had aWeb sire.

• Regarding the primary function of their company's Web site, 80% ofsurvey respondents indicated their sites were used for marketing orinformation, 60% reported their sites were used primanly for leadgeneration, and 55% reported their sites were used for sales ore-cornmerce,

• In 2000, among the companies wirh Internet sites, only 69% reportedmeasuring their effectiveness. For those companies measuring Website effectiveness, 77% measured the productivity of their sire by rhesales generated, 62% evaluated the effectiveness in terms of the leadsgenerated, and 60% used home page hits as the key measure ofeffectiveness.

When you establish your Web site, the first thing you want to do ismeasure its effectiveness to determine if it is meeting your requirements.Currently, there is not a single standard for measuring the effectiveness ofan e-commerce Web site. Traffic and revenue generated by a Web site can

Analyzing and Targeting Online Customers 335

be reported in a variery of ways: number of page hits, dollar sales, rimespentat the site, source of traffic, returning customers versus new visitors,andso on.The ways in which reports are generated requires a means of tracking

and managing the dara so thar ir can be extracted into reports Web trafficis usually tracked by rraffic analysis and monitoring sofrware, of whichvanous software packages are available. Most analyzers first compile thedara into a dara mart from which they can tun reports. Packages differ inthe process in which the dara mart is built. Some build the dara mart in realrime and others build ir by loading the log files during off-peak times.Some examples of various e-commerce reporting and analysis tools

that build dara marts and analyze informarion in real rime include rhe

following:

• Accrue Insight builds the database in real time and can collect andstore dara on more rhan 1 million hits per hour. The data are storedin rhe lowest level of detail, and the resulring database can providethe maximum number of options on sorting, analysis, and reportgeneration. This is a good choice for large commercial sites storedover multiple servers.

• Aria works in real time with separate components to monitor, record,and report the data. The monitor intercepts the data, the recorderprocesses the dara, and the reporter provides the dara in chart form.The dara are stored in an object-oriented database.

• Webtrends has the ability to monitor server statistics almost [rnmedi-arely. It uses background processing to gather the data it needs. Itdoesn't import log files into a database but instead creates reportson the fly as the files are being read in. If capturing log file data in adatabase for later analysis is of interest to you, Webtrends will not

provide a complete solution.

For a more complete report on available reporting and analysis tools,read Data Mining Your Website by Jesus Mena (1999). .

The purpose of reporting can be to measure the relative populanry ofvarious areas of an Internet site the popularity of specific pages to a par-ticular population, dollar sales ~enerated overall, or by a parricular popu-

lation) the proportion of repeat visircrs, and so on.For example, Ar Home Online, an e-railer of home-relared products,

rebuilt its Web sire in October 1998 to be more user friendly andinformarive. The following exhibits are examples of reports designed totrack relevant consumer information, by week) from their new Web site.

Exhibit 16.4 shows the total number of visitors and the number ofviSitors by specific pages (home page, the product seatch engille page, the

garden page, and rhe cooking page).

"'~.1:! .S *~-

""'00 00 00 '" '" -e- '" t-, '" 00 CO r-, '" 0000 co~ (l ~ '" '" '" '" '" r-, r-, r-, e-, " e-, r-, " '" '" '" '" '" '" "'''' "'0B c,

0

~V

.1:! "$ 0 1 0 0 0 '" 00 ; 00 -o M '" r-, '" M '" 0 a-: co 0 0" r-, " " '" '" ... ... M M M M '" '" '" '" '" '" '" "'B

0 ~~=~~~~~~~8~;~~~~~~~~~V g." ~~~~~~q~~~~~q~~~~~~~qq-3 c, ~ON~~~~~N--OO~WOO~~~~OOON~MNNNNNNNNNN _______ V~

~t V

.g 01:! .""$ £ 0 -o ... M 00 :J: M 0 0 M -o ... '" ... MUJ ; ; '" M '"... ... ... M ... -e- -e- ... ... ... ... ... -e ... ... ... ... ... ...B .c~

0V

"'

~B ~1:! d: -o "'''' r-, ..,

'" 0 M co '" M M -o '" -o -o '" M ..,'"$ '" "'''' 00 '" '" '" '" '" 00 a- '" a- a- a- a- a- a- a- a- a-a-VE0:I:

E ~o 0:I:"-

Analyzing and Targeting Online Cusromers 337

Exhibit 16.5 shows the dollar sales associated with specific product,ffmmes and overall sales generated by the Web site. The last column showsaveragesales per visitor.Exhibit 16.6 shows the time spent by visitors in specific areas of the Web

site as well as the total time spent by all visitors. The last column providesthe average time per visit.Exhibit 16.7 shows the net visitors by a referring source. The last column

reflects visitors who reached the Web site directly by typing the URL intotheir browser. This data could also be reflected as percentages.

Driving Customers to Your Web Site

Forsome traditional direct mail marketers, converting current customers toonline communications may not be easy. After all, a customer whoresponds to a mail order offer is different from one who responds to anonline offer. Despite the efficiencies to be gained in converting customersfrom traditional mail to e-mail, a loyal direct mail customer may resistdoing business over the Internet for many reasons:

• Technology. Doing business over the Internet requires, minimally, acomputer, a modem, and an e-mail account. Some markets, such asthe seniors' market, are lagging far behind the average in Internetpenetration.

• Preference. Some customers have a comfort level with doing businessby mail or over the phone and will never feel comfortable providingcredit card information over the Internet.

• Convenience. Some catalog shoppers make their selections whilewaiting for appointments at the doctor's office or at night in theirbeds. For these people, the computer does not lend itself to theconvenience of shopping when their time permits.

Still, the effort to identify the segment of customers sourced offline for con-version to online communication is worthwhile. The immediacyand efficiencvof the Internet medium permits the development of a personal relationship. I - "I keti .des marketersWit1 your most valuable custOmers. Ecmar mar enng provi

with the ability to increase conversion and retention rates by. being able to

deliver timely and powerful individualized messages ~o the :anous ~ustomersegments in ways that were not feasible with conventional direct m.aJl. .

F. . I . izing that mte!bgent

Or example, Fingerhut Compal1les nc., recogru J .i .. . I k b ildi g a successful busmess,nteractIon With customers IS rne ey to ui 10announced in April 2000 its plans to begin launching a series of targeted

"I ("F· rhut to Launch Newe-rna. messages to their catalog customers ingeE-Mail," 2000).

�~~~~~~~oo~~~~~~~~~~~~~dooooooooooooooooooo

M-.oM I.f'j~~ oo""'V-.ocO NcOo'..o"",,:;;:~.~~~~N~"""Lrl-.o -.oui'"'V 'V"t"I.f'j"it""it" v

~~~~~~~88~~M ..oOoo..oc'NcONcOg;.-.o~~~~R~~~O'Lrluiui-.iuiuiuiui...o..o",

~~O'cO ,....:"'''''"00. N L/'}O'r-f~

0'-.0 00 1"'1 MLrlMMO'Ne-~..ooN..oc'vicO~t!:.~~~.te,... ~00 1"'1 OON Lrloaf-N-NNM

co NM

<XiOM~~

N - N

~~~~"'~0<","''''

0"'''"'''''''"';..0,...;~LI"!.~...-"''""'''''''

"'OM"'''''''0"ci . ..f...-...-...-

0-.00' OM M .......LrlI.f)I"'I LrlOOO'..o'"'1v V.~LrlVOui""':oocO"';l.Iiaivvvv"t"U"lLrlL/'}

338

� -1! .S .:g 2 ~ -o o- N� C.~ 0 0 '" ": ~

-c 0 ~ O' ~~~ o- -o o-0 s: O' 0 0

0 '" co co -� E u -e- ~ ... ~~ ... '" o: 0 ui ui ui '" '" ""

~> >:: � - ... -e- ... ~ ... ~ e- ~~ "" '" co-c '" �

~~~~~~~.c,

---------

O~O~-~-~~M~~~OO~O-~~OO~~N~~OOOM-~OOOOO~NNO~M~OO~~~OO~-OOOON~~~~~~~~~~~~~~~~N6o~~ro~~N~VOOO~~~N~OOO~ro~N~~~~~~~~gg~~~~~~~~;~

..;g.c)(w

co co co co co cc cc co <Xl O' c-O' o- '" co a> O' O' o-: co O' U> O' U> U> U> O' U>

U> o- U> U> U> U> U> U> U> U> U> O' U> U> U> U> U> U> U> o-U> U> o- U> U> U> U> U> U> U> o-

- U> U>2; ~ - U>

U> o- U> '"-" '" 0- "' ;::;; go "' ;;; ;::: "'- - 00 Ui Ui N� a s ::: c:' '" '" "'- :::~ 00 '" ~~ 2; N N N N or - N N '" N M ~- - -

339

-i1s: t-, r-, cc .... co t- cc a; " -c t-, .... 0 ;;; "' M '" 0 a- M 0; Mco M cc .... o- "' -e- co -o -e- 0 '" 00 co e-, co 0

t ..... "! M• M. ..... "l "l "! -, r-, "'! "'. O. N M ~ "' -o cc-~ .... N N N N N N Na

~-sc:§ N "' 0 cc cc ;;; -o .... '" co '" '" cc "' '" 0 00 0 '" r-, M M

r-, .... co M '" N o- -o M '" r-, -o '" "' -e- '" '" -c <Xl 0, "! "1 "1 ..... ..... "l "'."! -, "'! "'. "'. 0 N M -e- '" <c " <Xl. 0~ N N N N N N N N ,.;-5 - - - - - - - - N

0

c,

'" M o- 0 -e- M co '" r--, N '" " '" '" 0 M N r-, o-: 0~ a; ;::-5 0 .... .... '" .... '" '" '" 0 " '" '" '" <Xl M M

"' <Xl - .... "'! .... "' '" .... <Xl. "1 '" N t- "1 <Xl• ..... 0 '" M 0 t-o '" '" as as <Xl '" '" '" 0 0 N N M M .... on on .,; '" '"~'"

00~ ~

'" '" - N 0 "' '" '" ;;: ... '" " '" ... N 0 00 '" '" <Xl N '"~ "" ... <Xl M <Xl ... 0 t- '" M M ... '" '" M <Xl M 0 <Xl " <Xl '"0 0 "1 .... "! " '" .... ... '" <Xl 0 N .... '" '" .... " '" N '" <Xl~ ~ M M M ,.; ,.; ..f ..f ..f ..f ..f on on '" .,; on .,; '" .,; .,; '" '" '""'E..g" ~ 0 M ;;; '" "' N "' ... '" " " '" M '" ... a- '" <Xl " oo~ 0 <Xl - '" ... 0 '" M '" <Xl <Xl '" M t- - " ... 0 0

U M ..... '" t- '" 0 N ... '" " '" ..... '" <Xl M '" '" "! "'" <l'. .::. ,.; ,.; M ,.; ,.; ,.; ..f ..f ..f ..f ..f ..f on '" on on .,; .,; .,; .,; " '"bO "'" 0:0-bOc:: " '" "' ;;: ... ... .... oo a; ... '" '" "' "' '" '" '" M 0 '" .... '" 0·C 0 t- - '" <Xl t- '" - ... <Xl M '" '" (1; ... '" '" 0 ...L ~ <Xl '" - M ..... '" oo 0 N '" " '" .... ... " M '" '" M "! °.2' ~ ,.; ,.; ..f ..f .... ..f ..f on on on on on .,; .,; .,; '" '" '" '" as <Xl '""'">-.D

:'l -5 "' '" '" <Xl ... <Xl M ... '" '" t- '" .... <Xl "' '" 0 <Xl 0 <Xl

'" .... 0 a- '" '" M '" 0 '" - <Xl t- '" t- a- M " ....·Vi 0 <ll M "' '" <Xl 0 M '" " O. N "' t- ~ M '" a- M '" 0 ..... oo'> 0 ..f ..f ..f ..f ..f .,; .,; on on '" .,; .,; .,; '" '" '" as as '" a- '"- is0 '"L ," ".D 0

E0 ~ '" N M <Xl <Xl "' t- " '" .... <Xl "' N N "' N 0 .... " '" ;;;0 0 "' '" ;:!: 0 <Xl oo '" N " ... N M '" <Xl a; 0 '" N N '"" <Xl

~ 0 M ... oo .... " "'. '" t- .... <Xl '" <Xl "' N '"~Z E as as as '" '" '" 0 0 N N M ,.; ..f ..f on -D -D '" as '"0 - - - -bO <tc::.~ "0 ~

" '" .... .... '" '" .... <Xl '" ... " M ... '" t- '" - '" N M '""" <l'. 0 t- '" .... '" M .... .... '" 0 t- " a- ;;: ... 0 M a- 0; ....Vl " ~ <Xl .... t- -. "'. '" " .... <Xl .... <Xl "' .... 0 <Xl '" ..... M "10: '" 0 0~ ~ N N M ,.; ..f ..f '" -D -D '" as '" '" 0 ;;; N M... - - - N N ....0"-"'"" .{'J 0 ;:: "' " '" ... " 0 " ... N " a- ....a. <Xl '" <Xl N N a- a-

S 0 M M .... '" '" " M '" " '" <Xl ;;; a- a- ... ... '" -E .... '" "'. '" 0 ... <Xl "! oo~~C'!. M "1 a- '" N '" -'" 'ItN ..f '" " '" N ..f .,; as M '" '" N on as ..f as N '" 0Jj ... .... ... ... ... "' '" '" '" '" '" '" '" " " t- oo <Xl <Xl '" a- °

.....,;<Xl <Xl <Xl <Xl <Xl <Xl <Xl <Xl <Xl... '" '" '" <Xl <Xl '" '" a- <Xl '" '" a- '" '" '" a- a-:c '" '" '" '" '" '" '" '" '" a- '" '" '" '" a- '" a- a- a- a- a- '"'" '" '" '" a- a- '" '" '" a- '" '" '" a- '":E "'" ;0 ;0 - 0 '"~~

M,., 05 '" a- a- a-~ g i':: 05 - ;n '")(

~ 0 0 <::' ~ N ;:; "' 05 iiiw '" '" - :!: <::' NN

'" '" '" '" ;;; M

340

Analyzing and Targeting Online Customers 341

One way to determine which customers are likely to be qualified onlinebuyers is to overlay enhancement data on your database regarding aperson's online behavior. But remember, just knowing that someone usesthe Internet is not sufficient, because not all online users are purchasers. Aswe previously stared, many Internet users are not buyers until they havebeen online for several years. You can employ regression modelingtechniques to determine those customers most likely to be wired andresponsive. Once identified, you can send offers inviting them to visit yourWeb site and register for a free gift or to enter a contest.The Reader's Digest Association recently tested sending a postcard to

their offline customers offering a free gift (camera or frequent flyer miles)in exchange for their e-mail address.

In June 2000, Experian and FloNetwotk announced a service that willallow clients to use the extensive information in the Experian database to tar-get Customers online. Providers such as AccuData America and DonnelleyMarketing offer assistance in straight appending of e-mail addresses (botbpersonal and business) onto a postal list, but the match rate and coverage arerelatively low. A good match rate ranges anywhere from 4% to 25%. Butremember, once the e-mail address is appended, it may not be yours to pro-mote until the names go through an opt-in campaign. Cbeck with the vendorsupplying the e-mail addresses. They may have already prequalified thenames, which means you can begin sending solicitations immediately.

You may also want to consider the use of Abacus e-Dirccr service to helpyou find which of your direct mail customers will be good online cus-tomers. Or you might consider InfoUSA's e-ShareForce database, which isan enhanced version of its long-standing ShareForce database, which nowincludes Internet and telephone response data.Retailers also are trying to encourage in-store shoppers to visit their Web

sites. CVSlPharmacy in May 2000 launched its first major online andoffline integrated promotion. CVS stores banded out cards sreenngCUStomersto their Web site with a chance to win a Palm Springs getawayvacation. Staples Inc. has also been highly promoting irs Web site bylaunching irs "Hey, you don't have to run out" campalg~l using m:,storesignage, its catalog. and srand-alone direct mail advettJsements ( CVS

Launches," 2000).

Targeting Online Customers

-r- . "'1 h ffline world and consists ofiargenng online customers IS very smu ar to teo -h k nd executing a contact strategy.two major steps: defining t e target mar ers a . hi I b hYour strategy dictates the offer and message and the mannerill W C.1 or ~re

d. a· Iy to acquire retain,

extended. Ewcommercedata can be use Just as ertecnve . .'. . tradirionaldlrecrmail.

reactivate and cross-sell online customers as

In this chapter, we discuss a number of marketing innovationsprecipitated by the Internet and the data acquired by these innovations.Customers interacting with a direct marketer's Web site or customers of apure play e-commerce site can be segmented on the basis of their attributesand their historical clickstream behavior.

Just as customers are segmented by demographics, source, productaffinity, or RFM for traditional direct marketing applications, they canbe similarly segmented by these attributes and more for e-cornmerceapplications. For example, consider our direct marketer ACME Direct. Acustomer who ordered offline recently and registered at the ACME DirectWeb site for order fulfillment information can be classified in a segment ofrecent ACME Direct orderers and also be segmented as an offline ordererwith a potential to convert to an online orderer,

For direct marketers conducting business both offline and online,segmentation analysis, as described in Chapter 8, will take on a newdimension. This is because the RFM data elements can be considered in thecontext of online or offline customer activity or both combined. Theappropriate definition of segments will be driven by the specific needs ofthe marketer and the strength of the data to predict customer behavior. Forexample, suppose a direct mail cataloger has just begun to operate in theonline world. Up to this point, the cataloger considered a primary directmail customer to be anyone who made a purchase in the past 18 months.Now that they are in the online world, should the cataloger also includeanyone in this primary direct mail marker definition who placed all ordervia the Web site too? The cataloger may decide to include some, but notall, online orders in the definition. They may decide to include only recentonline orders in their primary direct mail market definition if and only ifthe Customer had at least placed one direct mail catalog order in the past.In addition, they may decide to exclude recent online orders in their pri-mary direct mail market definition if the customer had never purchased viaa direct mail catalog promotion. As you can see, it becomes very tricky andrequires testing and analysis to determine the best way to segment the mar-ket. By not carefully planning the integration of both online and offlinedata in your CUStomersegments, you risk telling customers via an e-~ailmessage that you miss them, when in reality they just placed an offlineorder.

Web marketers also desire to maintain relationships when customersbecome inactive or if their product or service is associated wirh longpurchase intervals (e.g., high-ticket items such as home furnishings, cars, orappliances). Segmenting the CUstomer file based on the last Web site visitdate or last online purchase date allows direct marketers to effectively exe-cut~ a reacti~ation or retention program. And if they are also conduct~ngbusllless offlme, they will need to determine how to integrate any offlineCustomer activity into these segment definitions. By not doing so, once


Analyzing and Targeting Online Customers343

again, direct marketers risk telling a Customer via an e mail I. " -, - I message, natthey rmss them, when III reality they just placed an offline order.

You can even append enhancement data las described in Chapter 3) toyour online customer data (registration, behavior, and source). LookingGlass Inc. has developed 27 unique market segments describing onlinecustomers. These segments are similar to the modeled data described inChapter 3 but for the online world. Direct marketers can append such datato their file for purposes of more targeted messages and offers. Examples ofsome of their segments include

• "Jules and Roz": affluent couples with kids• "Kelvin": techie guys• "Alec and Elyse": empty nesters, average age 53, income> $100,000• "Jason": male students and graduates, average age 22, income

< $20,000

Segmentation allows marketers to maintain a personal relationship with(a) active Customers between purchases, (b) those who visited the site butnever ordered, (c) those who have been inactive for 6 months, and so on.Some marketers send their inactive customers e-rnails highlighting freeshipping offers or percentage discounts available within a specific windowof time. These offers may be sent via e-mail or traditional mail.

MotherNature.com segmented its file by those customers who registeredand ordered versus those who registered and did not order. In an attemptto convert nonordering registrants, MotherNature.com e-mailed thema special offer ("MotherNature.com Streamlines," 2000).For a consumer who is highly active over the Internet and browses and

shops at many e-cornrnerce sites throughout the year, these offers, thoughvaluable, may be ignored due to clutter in the e-mail box and a lack of dif-fetentiation from other marketers. Sometimes to reactivate a relationship, itpays to send a reminder via traditional mail. This is a strategy commonlyused by Amazon.corn.Ashford.com, the luxury jewelry e-tailer, invited its best customers in

1999 to a preview parry to view new, exclusive merchandise In advance ofthe 2000 holiday season. Customers meeting a dollar threshold of paymentsin 1999 were allowed to preview new collections prior to the newmerchandise being posted on the Ashford site. Invit~d ~lls:orners wereallowed a 20% discount on the new merchandise. The mvttatron was sentvia first class mail and required an RSVP. Receptions were held in severallarge metropolitan areas at exclusive hotels, at which Ashford.comCorporate officers were also in attendance.

It is doubtful that the attendees of the reception or even those whoRSvP'd in the affirmative were recorded on the Ashford.com database.G . . re a public relationsenerally, receptions to specific customer groups a

344 OPTIMAL DATABASE MARKETrNG

function, and no thought is given to updating the database with attendanceinformation. Thus it is impossible to answer the question, "How much dida customer's reception experience affect his or her purchasing behavior?"The database may have recorded which customers were invited to thereception, but in all likelihood it did nor capture those who attended. If acompany routinely stages special events such as the one described above,then it is imperative to capture attendance data to gauge the effect oncustomer behavior.

You can also define your online target markets with regression modeling.You can use Internet data to predict the customers who are likely to clickthrough and order via a unique e-mail offer JUSt as effectively as you canpredict order behavior fat YOUtamine customers. You will employ the sameregression modeling techniques learned in Chapter 10. The only differencewith online applications is the ability to isolate and identify key predictorsof information to use in the modeling process.The challenge for the analyst is to identify relevant variables in the

constant stream of information stored in the e-cornrnerce database. Factorand cluster analysis las described in Chapter 8) is of value to identifydimensions of behavior in data and to reduce many correlated variablesinto linear combinations of a few important variables.

However, in most cases, the application of modeling for the Internet willnot lie in identifying those most likely to respond, as is the case withtraditional direct mail. Given that the economics of messaging a customerelectronically are so low, the benefits of such modeling will never outweighthe Costs of building the model. But modeling can be used to identify theright prospect Or Customer for the right offer.

For example, with the use of regression modeling, neural networks,and other mathematical algorithms, online publishers and retailers aremaximizing their retention of Customers by delivering targeted and per-sonalized product offers, page content, and advertising based on real-timeonline Customer behavior data and individual preferences. Most softwarevendors offering such services to online publishers and retailers useproprietary technology for capturing and analyzing the customerinformation. Some software packages, when evaluating a customer, willexamine every Web page displayed, the sequencing of pages, and the timespent on each page.

A company like America Online also has a great opportunity to do thisevaluation because of the vast amount of Customer data they have on theirsubscriber base. All subscribers regularly receive various offers each timethey log On to AOL and during an active session. Ideally, these offers areuruqu- to each Customer's observed AOL behavior in addition to anyCustomer-provided information about likes and dislikes.In 2001, Predictive Networks developed an ad targeting system that

analyzes Internet users' keystrokes and mouse movements to differentiate

Analyzing and Targeting Online Customers345

quickly among household members using the same computer and account.Then they target ads to those individuals within the household.

Conducting Marketing Testsin the E-Commerce World

Just as was the case in the offline world, testing is also the foundation onwhich a direct marketer builds and grows an online business. When you arepreparing to implement an online marketing test program, you need toplan it properly to ensure that the results will be readable reliable andprojectable. Exactly the same appropriate testing principles should beemployed as is discussed in Chapter 14:

• Use of appropriate sample sizes• Observance of the "one change» rule• Leveraging test winners to the same universes in rollout

In addition, once testing results are final, you should assess those resultsusing the same tools as listed in Chapter 13: confidence intervals andhypothesis tests.

We next consider the two types of online tests for planning and analysispurposes: banner ads and e-rnails.

Banner Ads

Online banner ads can be priced on the basis of the number of impressions,number of orders or number of responses, depending on what you are, . .offering. An impression is the exposure of a banner ad to a Webpage visitor.Typically, direct marketers specify the number of impressions they desire. Oncethat number is reached the banner ad is pulled. A click-through on a bannerad is analogous to someone opening the envelope in traditionaldirectmail. Inboth cases, the individual is looking for more information about the offer..Those placements where your banner ads yield the I~oSt IInpresslol.1S

and the highest order conversion rates are considered winners. You willContinue to run banner ads on those Web sites.For example, Vita-Protein wants to test several banner placements to

determine which one provides the best click-through rate for a newwomen's hair care product. The banner ad will highlight a free sample and

I· k h b ad they will be raken rocoupon offer. When prospects C rc on t e anner, . .. ddr and e-mail address IIIa page where they provide their name, a ress, .

. I ail Vira-Prorein haseXchange for a product sample and a coupon in rne m I..

d id d on Exci ·V·llage and Drugstore.coII1eCI ed to test this banner a on xcrre, I l ,


The marketing director at Vita-Protein, with the assistance of his mediaplanner, determined that the expected click-through rates for Excite,iVillage, and Drugstore.com will be 0.003, 0.0045, and 0.004, respectively.Assuming that the marketing director wants the click-through rates receivedfrom the Excite, iVillage, and Drugstore.com tests to be within 3% of theactual click-through rate to expect in rollout with 95% confidence, howmany impressions does he need to order for each site?

Using the sample size formula from Chapter 14 (Sample SizeDetermination for the Difference Between Two Sample Proportions), themarketing director set up the Excel spteadsheet as shown in Exhibit 16.8.

To gauge click-throughs at this level of accuracy would require18,247,600 imptessions of the Excite ad and would potentially yield52,918 to 56,568 click-throughs.

Assume the marketing director contracts through his media buyer for18.3 million impressions 011 Excite, 10.1 million impressions on iVillage,and 12.7 million impressions 011 Drugstore.corn. The order went into effecton April 14,2000. At the end of 3 days, he receives the report shown inExhibit 16.9 from his media buyer.

Should the marketing director adjust the number of impressionscontracted to ensure that he maintains the desired reliability in test results?It appears that the click-through rates are on forecast and that noadjustment to the number of impressions required is necessary.

Exhibit 16.8 Sample Size Determination Calculations

Error Rate as Sample Size Expected Lower UpperEstimated a Percentage Required # o( Bound BoundClick- Acceptable o( Click- (or 95% Click- (or # (or #through Error through Confidence throughs o( Click- o(Click-Placement Rate Rate Rote (%) Level @95% throughs throughs

Excite 0.003 0.0001 3 18,247.600 54.743 52,918 56,568iVillage 0.0045 0.000135 3 10,012.401 45.056 43,704 46,407Drugstore.com 0.004 0.00012 3 12.671,944 50,688 49,167 52,208

Exhibit 16.9 Click-Through Summary Report

Click- PercentageClick- Impressions throughs as Click- Percentagethrough Contract as of as of throughs of of ImpressionsPlacement Rate Impressions 4116/00 4116/00 4116/00 DeliveredExcite 0.00293 18,300,000 3.244.018 9,505 0.28 17.73iVillage 0.00446 10,I00.000 1,189,473 5,305 0.44 11.78Drugstore.com 0.00389 12,700,000 1,094.856 4,259 0.40 8.62

Analyzing and Targering Online Customers347

Once the tesr resulrs are final, the marketing director will compare theresults via hypothesis resring (as described in Chapter 13) to determinewhich were the winning sites.

Ir should be noted, however, thar the banner ad cosr model is changingdue to the slowdown in the Internet business world. It is now becomingmore common for search engines and Web sites to offer banner adplacement to other e-cornmerce businesses on a per response or order basisversus on a per impression basis. In other words, it is becoming much morecompetitive. Direct marketers 110 longer want to pay for impressions;instead, they want to only pay for responses or orders.

A new benefit being offered by many media buyers in the Internet arena isthe capability, via optimization software, to shift media dollars on the fly,based on rhe performance of aU media in a campaign. Consider an exampleof a product targeted to women: The media plan is to run restand control adson women's interest sires and general interest sites. If the test ads on the gen-eral interest sites seem to be receiving more click-throughs, the optim..izationSoftwarewould alert the media buyer to pull advertisingon the women's sitesand add more exposures for the test ads on the generaJinterestsites.

If your media buyers use such sofrware, you would be well advised to askthem the following quesrions:

• Are decisions to pull ads based on statistical significance?• If so, ar what level of confidence are the rests conducted?• Is the level of confidence appropriate, given the specifics of rhe offeradvertised and the associated profit impact?

E-Mail

D .. . . .. I to designing direct maileSlgnmg e-mail marketing campaigns IS ana ogouscampaigns. To ensure a certain level of confidence in your test results, applyth . .. Ch 13 d 14 to all e-mail Jist, offer,e appropnate rules given 111 apters an <

and COpy tests.. I· I· that the rules areWhen you conduct e-mail response isr tests, rea rze "J

different from traditional direct mail response list rests. In e-mail. . id d rhe owner for approva .marketing, a list is rented and copy IS provr e to :

Th d desi ted by the list owner or ae execution takes place by a ven or esignathird-party vendor who conducts broadcast e-mails. Some arrangelmel~ts" . endorsement by tne isrInvolve a stipulation that the copy contains an I h

. h maximum message engtOWner.In other cases, the list owner as a Ik d't bother to merge purge,requirement. Many e-mail mar erers 011 " d

b " . . I J access to their names anecause some list owners ngldly contra rne c h . divid I receivesh . . . that eac 111 IVI ua

t ere is little monetary benefit In ensuflllgonly One message.


The various arrangements that list owners require to communicatewith their names may make the management of a test mate difficult.However, as long as the responses ate ttackable by each specifictreatment, marketers should be able to cope with the new rules. In fact,the rewards to marketers are great if they can get e-mail marketing towork for them. The costs are lower, and the response time is immediate.With proper test planning and analysis, online marketers ace in a strongposition to make the best decisions regarding offers, treatments, andcommunications.

Chapter Summary

Many of the data mining techniques discussed in previous chapters can alsobe applied to online databases. However, some unique aspects of onlinedatabases need to be treated differently, In addition to customer contactand demographic and psychographic data collected during registration,marketers can track customer behavior on the Web site. This informationcan be valuable for customizing the site to the customer and also deliver-ing more targeted offers. Because there are cost and other benefits of doingbusiness on the Internet, organizations are attempting to drive theircustomers online. Although not all customers will convert to onlinepurchasing, database enhancement and modeling techniques may be usefulin segmenting those Customers most likely to convert.

Review Questions

1. What are same of the types of data collected from online customersthat are different from data collected from offline customers?

2. Discuss the characteristics of online buyers.

3. How is Web site effectiveness measured, and which of these meas-ures are most important to online marketers?

4. Discuss the advantages of real-time databases.

5. Discuss the process of online testing.

6. Why are marketers attempting to move their customers online, andwhat are Some of the methods marketers are using to do so?

Issues in the MarketingEnvironment and

Futu re Trends InMarketing Databases

The second issue-developing policies for data acquisition and databaserental-is easier. One thing she has to decide is whether to have customersopt-in or opt-out on renting or trading their names to other marketers.With the opt-in policy, customers had to specifically indicate that they wantto receive information from other marketers; otherwise. Inside Sourcewould not sell the customer's name. With the opt-out policy, Inside Sourcewould be able to sell names to other marketers unless their customers indi-

cated that they did not want this to happen.

In Keri L 'I ...ee 5 nez,v YO e as corporate Vice president of inside Source herdecisions are more strategic in nature and undoubtedly have a IOHger:termeffect On the organization Her decisions in her previous position mighttnuolue segmenting the database to select names for a mailing or testing anew promotion. These decisions were short term. III her new position, shealso is responsible for strategic decisions such as deciding to enter new mar-kets or setting policies that would have a long-term effect on the database.

These decisions also include expallding into global markets and establish-mg policies for the acquisition and rental of databases. Ken had alreadystarted to examine internatianalmarkets and saw au.immediate potentialfor her product in Europe and a growing potential in other global markets.It was not a question of whether to expand into global markets-it was aquestion of how and when. She II.OWhad to develop strategies and tacticsfor moving into global markets. Much more information was needed onculture, language translations, local competition, distribu.tion channels andlaws, and especially laws related to "direct marketing. She realized thatother factors would also come into playas she moved through the processof introducing the product into global markets. Keri anticipated that shewould need to include consultants' fees in the budget, because the company

Was inexperienced in international marketing.

350 OPTIMAL DATABASE MARKETrNG

For e-mail marketing, Keri was almost certain that she would implement astrict opt-in policy. There was a growing movement, supported by severalbusiness and consumer groups, to eliminate unsolicited commercial e-mail.Personally, Keri did not like receiving commercial e-mails unless she specif-ically requested them. She knew that broadcast e-mail was relatively cheap,but she did not want the future of e-commerce to be a battle by consumersof filtering through possibly hundreds of e-mail messages. Although unso-licited broadcast e-mail might yield profits for the company, Keri wasconcerned about the social implications of this action.

Keri was leaning in the opposite direction on, a policy for regular mail.Inside Source had purchased lists in the past from a number of sources, andKeri was sure that very few of the customers on those lists opted-in. InsideSource also rented lists of their customers. Inside Source's current policywas to assume that the customer's name could be rented unless the cus-tomer made a specific request to opt-out of other mailings. Even thoughsome public interest groups are opposed to unsolicited traditional mail,there are not as many strong feelings against it as there is against unsolicitede-mail. Personally, Keri did not mind receiving unsolicited offers in the mail.Before making a final decision, Keri was going to again review informationfrom consumer advocacy groups and industry organizations.

Marketers. have [~ c~n~ider more ~ha~lpotential profitability w~endeveloping, mall1tall1lJ1g, and utilizing databases. Global, SOCIal,

and legal factors also need to be considered. Larger companies have longused databases for marketing to customers on a global level. Morerecently, smaller companies have been able to use databases to facilitateglobal marketing. Factors like increased accessibility to technology andthe Internet have thrust these smaller organizations into the global arena.Although the implications of global marketing are complex, in thischapter we examine SOme of the factors that are particularly relevant todatabase marketers.

Database marketing has come under close public scrutiny in recent years.Public interest groups and governmental agencies have examined the data-base policies used by various organizations. This has come about becausethe growth of Internet commerce has raised concerns over how customerinformarion is accessed and used. The security of databases is a concernsparked by the unauthorized access and use of personal information andcredit card numbers. Traditional direct marketers who use mail and phonesolicitations have also Come under increased scrutiny. In particular,marketers who use sweepstakes promotions and target older consumers orchildren have become the focus of public advocacy, legislation, andlitigation, In response to potential legislation, many in the industry havecalled for self-regulation. The Direct Marketing Association (DMA) has

351

established a Privacy Promise for its membership In rhi I. . . IS C 13pter, we

exam me pnvacy concerns from different perspectives.The final section of this chapter is dedicated to future trends in marketing

databases '. For example, it is apparent that marketing on the Internet willshow significant growth in both consumer and b-to-b markets. Because themarketing database is a fundamental element of internet commerce we exam-ine the evolving role of the database in Internet marketing. In addition, data-base marketing has been expanding in a number of different business areas.Wetherefore examine the challenges database marketers face in these areas.

The Global Business Environment

As markets become saturated in the United States, companies must look tothe global business environment and evaluate markets in other countries.Continued growth is dependent on global expansion, and for many organ-izations, significant revenues often come from multinational operations.For database marketers, global marketing means a database system thatis integrated and synchronized across all countries in which companiesoperate. Dell Computer Corporation, for example, markets productsdirectly to consumers and businesses in 170 countries and maintains salesoffices in 34 countries. More than 30% of Deil's revenues come from otherCountries. Because 40% to 50% of sales are Web enabled, Dell requires anintegrated database system that transcends national borders (Dell, 2001).

Inter-Continental Hotels and Resorts also uses a database that goesbeyond U.S. borders. Inter-Continental tracks customer information frommore than 100 hotels located in more than 60 countries. In an effort tobuild customer loyalty, Inter~Continental built a central data warehouse.One of the problems they have is matching data from customers as the dataenter the database from different locations. For example, ill an English-speaking location, a hotel employee would enter the home address of a pa~-ticular customer as "London, England." However, the same customer saddress entered in a hotel in a French-speaking country might be "Lundres,Angleterre." Because of the cultural/language factors, Inter-Continentalrelies heavily on numeric identifiers like credit card numbers and phonenumbers to match guests correctly ("You say London," 1997). Inter-Continental currently has several programs that use the customer darabaseto build relationships, including travel mileage and acc~mmodanonupgrade programs (see www.interconti.com). These relarionsh,p programsalso transcend U.S. borders. Database marketers face unique challenges 111

global markets. In particular, the legal, poljtkal, and culrural envir~nmentsin other countries may differ drastically in the area of consumer pnvacy. Inmany cases, the organization must radically change the way they do

business as they move into other countries.


Exhibit 17.1 shows some of the factors that challenge databasemarketers in the global business environment.

We present a brief overview of each of these factors.Culture includes values, customs, rituals, symbols, roles, language, and

other aspects of society that are generally accepted by people in a country.The database marketer has to consider several elements of culture. On ageneral level, the marketer has to consider how consumers with a particularcultural background perceive database marketing. Privacy is more highlyguarded in some countries; for example, countries of the European Union(EU) have more restrictions on how consumer data are used by organizations.Numerous nuances of culture can affect database marketers. Latin

Americans, for example, may be more serious and formal than people fromother cultures, particularly when it comes to work. Therefore, using humoror cartoons in direct marketing communications to Latinos has to be donecarefully or not at all. But with Latinos it could be beneficial to focus onstatus. A mailing that looks official or expensive is likely to receive moreattention (Haegele, 2000).

Because some Asians feel that debt is a disgrace, credit card marketershave to be careful to communicate properly to them. Positioning the card

[ Political~

Culture ! I Legal~

Demographic i I PsychographicSocial

~ I ORGANIZATION ! Media~

Economic!I Competitive !Infrastructure

IHuman Resources!Technology

Natural Environment

Exhibit 17.1 Forces in the Global Business Environment

Marketing Environment and Marketing Database Trends 353

as being safer .th.a~carrying cash or for use in emergency situations mayovercome. the 1Il1t1a1negative attitudes. Because Asians in general like toread,. a direct mail campaign might be appropriate. In addition, buildingrelationships with ASians through multiple contacts is important.Marketers also have to be careful about overgeneralizarions. Just as not ailLatin Americans are alike, neither are Asians. The Chinese from HongKong are more savvy when it comes to credit cards than people from majorcures II1 the People's Republic (Shermach, 1998).Marketers also should consider whether shopping customs are consistent

with direct channels that database marketers use. U.S. consumers have longembraced catalog shopping and are finding that shopping on the Internet is.a logical transition. Credit cards are the common denominator in thesetransactions. Many Asian consumers~ on the other hand, deal in cash sales,are less familiar with catalog shopping, and are worried about potentialproblems with credit card transactions. Furthermore, the social interactionof shopping, especially haggling, is more important in many countries thanit is in the United States. Therefore, database marketers, who depend oncredit card transactions and not face-to·face selling, have particularchallenges when moving inro some Asian countries (Schmit, 1999).The social and cultural envircnrnenrs are interrelated. Societies have

varying degrees of demarcation for individuals on the basis of heredity,wealth, education, occupation, and other factors. In some societies, socialclasses may be clearly defined, and social mobiliry (moving to other socialclasses) may be difficult because of prescribed class standing. Furthermore,it may be inappropriate for an individual to interact with people of a dif-ferent social class. The roles of men, women, and children also vary fromsociety to society, along with the influence of family members in purchasedecision making. Therefore, it might be improper to target females incertain countries with offers for products that are considered primarilythe responsibility of males (e.g., financial services and insurance).

The legal environment includes the laws and legal practices of a countryat several levels-countrywide, state or regional, municipal, and religious.In addition economic communities such as the EU have laws and regula-tions that affect database marketing in several European countries. At avery basic level database marketers must comply with all relevant laws. InMexico, for example, rental lists are sometimes acquired illegally fromvoter registration rolls and income tax records. Marketers must be sure thatany list they use was obtained legally and that they have the right to use It

(Sutter & Mandel-Campbell, 1998). . .At another level database marketers must be careful to aVOid practices

d.. ' . I·· I . oups and legislatures to

an policies that might move po inca action gr .enact laws. The EU's Privacy Directive is currently a pnmary concern ofmany U.S. database marketers. The directive stipulates rhat companies may. .. tries that do not havenot transfer data from European CItizensmro COWl

354 OPTIMAL DATABASE MARKETfNG

privacy laws that are compatible with EU laws. The United States iscurrently not considered a safe harbor for data from customers in EDcountries. Individual U.S. database marketers ate hoping to negotiate safeharbor status with the EU pending their conformance to EU pnvacystandards.

One of the fundamental premises of EU privacy laws is 10 acquirespecific expressed permission from consumers before data are used forother purposes. The favored policy in the EU is opt-in. In the United States,most marketers follow an opt-out policy, and it is assumed that marketerscan use customer data for marketing purposes (e.g., rent a customer's nameto another company) unless the customers request that their names nor begiven to other organizations. With the current political and social environ-ment in the United States leaning toward more consumer control ofpersonal data, some people in the industry are calling for databasemarketers to move to an opt-in policy before laws are enacted and there isno choice. (Orr, 2000).

In some countries, laws restrict access to consumer information that isimportant for efficient database marketing. In the United Kingdom, forexample, access 10 the electoral roll is restricted. This roll has beenvaluable to direct marketers as a data-cleaning tool, because it containsupdated records of personal details. Without access to this database,direct marketers have difficulty keeping databases up to date(Gordon, 2001).

Direct marketers must also be aware of legal processes and proceduresespecially as they relate to gray areas in the law. A foreign organization maynot have the same flexibility in certain legal processes as would a domesticorganization. This often means obtaining legal representation or consult-ation in the country of concern.

Demographics includes a number of variables that we have discussedthroughout this book such as gender, age, income, family status, and resi-dence. As in the domestic market, rhe database marketer has to determinethe fit between a country's demographics and potential market offerings. InCOntrastto other countries, the United States has a population with a higherpercentage of older people. In some developing countries, most of the peo-ple are under 20 years of age, but in the United States there are more peopleover the age of 35 than under 35. The database marketer should be awarethat nonsegmenred demographics may be misleading. For example,countries such as China, Brazil, and India with low median incomes mayhave sizable affluent segments that may be potential customers for higher-priced products.

Although culture refers to generally accepted values, customs, rituals,norms, and so on, psychographies refers to activities, interests, and opinions(AJO) of segments of a country's population. Psychographies may be moreImportant than culture or demographics in evaluating marketing potential.


III

The AlO of a segment of a society may be in conflict with culture. Forexa~ple, nonmarerialisrn may be a cultural component of a country such asIndia. However, as we just mentioned, sizable segments of the populationmay be more oriented to Western consumerism and luxury products.Database marketers should consider the potential of these segments but mayhave to adjust the positioning and pricing of luxury items consistent with theculture (Mistry, 1996).There are obvious implications of activities and interest variables. A

database marketer who offers gardening tools and supplies should look atthe number of people in a country who report that interest. AIO may bederived from a number of sources. For example, if direct reports of interestin a particular area are not available for a particular country, the marketermay be able to estimate interest based on membership in interest groups orsubscriptions to publications in the interest area.As with the domestic market, evaluation of the competitive environment

in global markets is essential. Does an existing company currently have afirm hold on a market in a foreign country that you are considering? Forexample, in Germany, the retailer GalerialKaufhof is a division of MetroAG, one of the world's largest retailers. GalerialKaufhof sells directly toconsumers and through brick and mortar locations. Their productassortment is very broad, ranging from wine ro watches. Direct marketersconsidering selling watches in Germany would have to evaluate thechallenges that a competitor like Galeria/KauJhof would impose on theirmarketing efforts. On a strategic level, marketers have to consider factorssuch as company resources, product assortment, target markets, position-ing, reputation, and customer loyalty of the competitors. On the tacticallevel, a native competitor may have strong relationships not only withCustomers but also with suppliers. For direct marketers, establishingrelationships with list brokers, mail shops, fulfillment centers, and othersuppliers is important to operate efficiently. Native competitors often havean advantage in these areas. . .The economic environment includes a number of elements, I11cludmg

employment levels cost of living, exchange rates, inflation, and poverty, " dlevels. The infrastructure component encompasses commulllcatJons antransportation. A country with economic data that is inconsistent withdomestic economic data (e.g., extent of infrastructure development) maystill be a reasonable candidate for market entry. However, databasemarketers need to make approprjate adjustments (e.g., alternative.distribu-. hi J k t Fo example adJusonentstton systems) when approac ing tnese mar e s. r e: ,need to be made if transportation channels are inadequate. Some more

, ' kaai h or may not be able to beperishable items may require pac aging c anges -sold at all. In the United States, a responsive djstribution syste~ ~anmove

, f h 'S 'f necessary This IS not rheproducts across the country In a matter 0 om [ .case in many developing countries.


Political environments and forms of governments vary widely. In particu-lar, political instability (e.g., Pakistan, Russia) can offer an unknownelement to the marketer. The legal and political environments areinterrelated. Although certain pracrices may be rechnically legal in acountry, government policies may make them impractical to implement(e.g., obtaining appropriate permits). Direct investment in a country, ascompared to less risky operations such as exporting, should be evaluated interms of political and economic stability. Setting up database facilitieswithin a country can constitute a substantial investment. The databasemarketer must evaluate how current and potential future governmentsmight react to the organization's initiative in the country.Human resources include employee expertise, employee recruitment, and

employment policies. Database marketing requires technical and marketingexpertise. If these experts do not exist in the country, the organizationsmust consider alternatives such as employee relocation, training, or operat-ing the business from another location. Companies like Allstate Insurancehave established international technology centers in Ireland because of rheavailable pool of well-educated programmers and computer technicians("Allstate Unit," 1998).The technological environment involves all aspects of technology (hard-

ware, software, communication links, peripherals, etc.) related to databasedevelopment and maintenance. Human resources are also involved in thetechnology area, because trained professionals are an essential componentof database technology. The cornerstone of direct/interactive marketing isthe computerized database. If access to database technology is limitedwithin the country of concern, then the organization will need to makeadaptations. Competent servicing, repair, and maintenance within the coun-try are necessary. If they are not available, the organization must makeadaptations. Sometimes the adaptation means housing the technology inanother country and using only minimum services in the target country.Although the natural environment has less of an effect on businesses in

developed countries, it may have a significant impact on businessesin lesser-developed countries. Weather conditions, topography, pollutionlevels, and so on can affecr business. Technology may interact with thenatural environment and infrastructure. Sophisticated computer systemsoften require controlled environments. If weather conditions are extreme(e.g., high temperature, humidity), the environmental control systems mustbe reliable enough to avoid major system failures. In addition, the markerermust determine if media and distribution channels are susceptible to

seasonal disruptions (hurricanes, floods, snow, etc.).Media for database marketers are the means of communication such as

mail, print, telephone, TV, radio, and the Internet. In some developingCountries, mail delivery is inefficient and telephone service may beinconsistent. Even if mail and phone services are adequate, it may be diffi-

Social Concerns andEthics in Database Marketing

Marketing Environment and Marketing Database Trends

cult to find appropriate lists. Database marketing is growing significantlyIII Brazil, but direct marketers face several challenges. The quality of the~ames ,on list rentals IS poor, mail service is inefficient, and phone service isinconsrsrenr (Sutter & Mandel-Campbell, 1998). Despite these problems,companies like Lloyds and Ciribank invest in direct marketing programs inBrazil. These cornparues believe that the risk is outweighed by the strongporential growrh (Molloy, 1997) .. Culrural and lifestyle characteristics interact with media. In the United

Kingdom, for example, a majority of the population read national daily andSunday papers. The weekly reach for all commercial radio there is 60% ofthe population. Free-standing inserts are popular in Europe and Japan. Radioand TV are significant media channels in Germany (Yorgey, 1998). The keyquesnons regarding media and global marketing involve the appropriatenessof available media for reaching rargets, the organization's compatibility withthe media, and how the database interfaces with the media.

Database marketers must be concerned with how their actions affectvarious segments of society. These segments include not only customers butalso prospective customers, public interest groups, regulatory bodies, andlawmakers. There are long~te[m implications if the actions of organizationselicit a negative response from these segments. For example, the way inwhich some companies used sweepstakes promotions elicited a negativepublic reaction and legislarion that affected the entire industry.

Exhibit 17.2 shows areas of concern for database marketers.At the base of rhe pyramid, marketer have more concrete guidelines for

marketing actions. For example, according to FCC regulations, there is aban on the use of fax machines to send an ;'unsolicited advertisement" (47USc. B 227 (b)(l)(C); 47 C.ER. B 64.1200(a)(3)). If direct marketersviolate laws, the penalties are a clear motivation to keep company practices

wirhin the legal boundaries.As we move up the pyramid, the motivation for avoiding or imple~en~-

ing a particular practice becomes less clear. When the political situatIOn IS

prone to enacting legislation, direct marketers are often motivated to

change their practices. Most of the time, an industry organization such asthe DMA spearheads efforts to have rhe industry regulate itself before lawsare enacted. The DMA's Privacy Promise, which we present later 10 ~IS

chapter, is a self-regulation effort. Certain practices can have negat~veeffects on some segments of society. These segments ~ay not have the SizeOr political power to directly affect the political enVIronment. However,Over time, public policy groupS may adopt the cause of this segmenr.

357


Social

Political

Legal! Regulatory

Exhibit 17.2 Areas of Concern for Database Marketers

Children, for example, don't have direct political power, so children'sgroups have adopted the cause of protecting children from certain types ofdatabase marketing.

Legislative action may have been taken on sweepstakes promotionbecause of the political power of a social group. Older citizens have highervoting rates than other groups, and therefore legislators are more likely toheed their concerns. The American Association of Retired Persons (AARP)has taken on the cause of protecting its members from fraudulent market-ing practices and viewed certain sweepstakes practices as a major problem.Legislarors heard testimony from directors of AARP and evenrually enactedlaws that are intended to protect older citizens and others from practicesthat were interpreted to be deceptive or misleading.The Honesty in Sweepstakes Act includes the following provisions:

• There must be clear and conspicuous statements in the mailing, rules,and order forms indicating that no purchase is necessary to enter thesweepstakes and a purchase will not increase the odds of winning.

• The mailing must not claim that someone is a w.inner unless theindividual is actually a winner.

• The rules must contain a description of the prizes and the odds ofwinning.

• An address or toll-free number must be included in the sweepstakesmailing, allowing a recipient or their assignees to prevent future mailsweepstakes promotions.

F

Marketing Environrnenr and Marketing Database Trends--------~ 359

In an effort to strengthen relationships with consumers, governmentalagencies, and public interest gtOUps, the DMA established a SweepstakesHelpline. The purpose of the Helpline is to educate consumers aboutlegitimate versu fraudulent sweep takes and offer assistance in directingcomplaints and name removal requests to marketers. From the perspectiveof database maintenance, these laws require a higher level of vigilance toensure that con urner who do not wish to receive mailings arc properlyflagged in the database.It might be beneficial in the long term to become responsive to the

feedback from social groups before the issue moves into the political arena.This is a much more difficult task for the direct marketer, because theconcerns of these groups have not been clearly represented. On the top ofthe pyramid are ethical concerns. Marketing practices exist that might notelicit social, political, or regulatory actions, bur these practices might bequestionable from another perspective. For example, direct marketers havemade mail pieces look like official government documents. Although thereis nothing illegal about this practice and it is not currently eliciting a largepublic reaction, some people think rhe practice is deceptive. Consumersopen the mail because it looks like an important document, and when theyrealize it is an offer for a product, some consumers become angry. Thesenegative sentiments do not help the image of rhe industry.As times change, issues can shift downward on the pyramid, For

example, questionable (deceptive) sweepstakes techniques that were onceused with little controversy evolved through social, political, and legalarenas. These techniques are now susceptible to legal action. . .

TI. . keri plex Most legItimate

ie Issue of ethics in direct mar eung 15 com ." . I d t f "auduJenrly take money

crgamzarions emphatically state that (ley 0 no I. .,'

f F 'stance IS It ethical torom COnsumers but there are some gray areas, or 111 , .

d.' I' S a check or IS from a

eceive people into thinking that an enve ope contam . ff .)" I I pe and review the 0 .errgovernmental agency so they will open rne enve 0 . ' . bI. . ' . I rl eir subscnptlon IS a outs It ethical to give consumers the lInpresslOn that 1 d h ld I,I renewal an 0 t ieto expire so that the company can get an ear y. . hi I [0 use

. . . ings on it? is It et caconsumer's money longer to gam Interest earru " . I te, b t mptmg to malupu aa contrived "survey" to solicit donations y at e '1 Hers to, . " hi I to send e-maJ 0attitudes with biased questions? Is It et rca h tS to remove. . . . t for t e prospecprospects and make it difficult or mconve",en . I d by directh. "Th P 'aC[Ices t iar are use

t err names from the mailing list? ese are I .

k Id -zue are unethical.mar eters that some people wou arg , d "by a particularI

b onsldere negativen addition, something that may e c , ty For example,

d I rhrea t to SOCle .group would not be considere a genera d ggregate rather thanb if " "often base on a .ecause credit risk classi IcaDon 15 . ff people in certain Zipi di . . k r rs restOct 0 ers to . I"n ividual data some direct mar e e h h is a higher like I-d

' , d indicates t at t ere ICo es. An analysis of the It house ara d f I payment. Therefore,h " d will e all t on cood that people within the Zip co e


these people are not given the same credit terms that people in other areasreceive. Although this practice makes good financial sense for the business,some people believe the practice is discriminatory, because "bad risk" zipcode areas can have a high proportion of certain minority groups.Furthermore, an individual consumer in a bad risk zip code with a goodcredit rating will be lumped in with others. Some would argue that thispractice is unfair.

Industry Organizations

Industry organizations perform important functions within an industry. Forexample, the DMA helps in these ways:

• Represents the interests of industry members to the public and togovernmental agencies

A key goal is to promote self-regulation through the establishmentof guidelines for accepted practices.

• Promotes education in the industry by sponsoring conferences andworkshops

The Direct Marketing Educational Foundation specifically servicescollege students and professors.

• Conducts research On the state of the industry.• Acts as an intermediary for the collection of names of consumers who

do not wish to be contacted by direct marketers.• The DMA's Privacy Promise (see Exhibit 17.3) was developed as a

means to promote the adherence of members to certain privacypractices and applies to both consumer marketers and suppliers.

Industry organizations are concerned about privacy protection for agood reason. A study conducted by Voter/Consumer Research for theAssociation for Competitive Technology (Bremner, 2001) indicated that76% of the consumers polled said privacy protection is a priority withthem. Relative to other social issues, privacy protection ranked seventh outof nine in the following group: education, crime, health care, energy, socialsecurity, Medicare, privacy protection, national defense, tax cuts.. Without industry initiatives, legislation is more likely, but in some cases,industry initiatives are not considered adequate. For example, Congressis considering more legislation to protect the privacy of students; thelegislation would have an impact on direct marketers who target familieswith school-age children (Campbell, 2001).

Another industry organization, the Electronic Retailing Association(ERA), has set guidelines for telemarketing. Members of the ERA, includingorganizations such as ]. C. Penney and QVC, are asked to adhere to


Exhibit 17.3 Direct Marketing Association Privacy Promise

Privacy Promise Member Compliance Guide

Consumer Marketer's PromiseI cerrify that my company:(please initial)Provides custometS with notice of their ability to opt out of information rental sale

or exchange. 'Honors customer opt-out requests not to have their contact information transferred

to others for marketing purposes. -Honors consumer requests for in-house suppress to stop receiving solicitations from

our company. _Uses The OMA Preference Service suppression files, which now exist for mail, tele-

phone lists, and e-mail lists.

Supplier's PromiseI certify that my company:(please initial)Encourages our Consumer Matketer customers to comply with the DMA Privacy

Promise. _

Business to Business, Resident/Occupant, InternationalI certify that my company is exempted from the Privacy promise. -This certifies that the below named company is in full compliance with the PrivacyPromise, as described in the Privacy Promise Member Compliance Guide, receipt ofwhich is hereby acknowledged. By my signarute, I certify that I have personallyreviewed the company's practices that are subject to the Privacy Promise, and thatI am my company's designated contact authorized ro make this certification of

compliance on behalf of the below named company.

Source: © Direct Marketing Association. Used by permission.

guidelines on billing practices, disclosure of information to consumers,privacy, and the use of certain communication technologies.

Sometimes industry organizations advocate different positions on issues.For example, the Internet Direct Marketing Bureau (IDMB) developedguidelines for e-mail marketing that identify spam as being rude andirresponsible. 10MB endorsed a strict opt-in policy for compiling mailinglists. For example, a consumer who has recently made a purchase on theInternet would have to check a box on a registration form before marketerscould send marketing information to the consumer or share the consumer's

e-mail address with other marketers.The OMA on the other hand currently advocates an opt-our policy for

, 'e-mail marketing. However, all DMA members who wish to sendunsolicited commercial e-mail are required to purge their e-mail lists of


individuals who have registered with a DMA service called rhe E-MailPreference Service (EMPS). EMPS purs individuals on a lisr if rhey do norwish to receive unsolicited commercial e-mail.Industry organizations in various areas of database marketing are facing

substantial challenges in light of public and legislative attitudes towardconsumer privacy and marketing fraud. On one hand, these organizationsdo not want to place unreasonable burdens on the businesses of theirmembers by establishing policies or guidelines that require substantialresources or restrict productive business practices. On the other hand, if thepublic and legislatures do nor feel rhar rhe self-regulation efforts aresufficient, the possibility of more stringent legislation increases.Exhibit 17.4 is a letter ro Congress from NetCoalirion, an industry

organization that seeks self-regulation of the internet in the areas ofprivacy and database protection.As you can see from the letter, if legislation is necessary, NetCoalition is

seeking narrow versus broad approaches to strike a balance between publicconcerns and restrictions on companies in the industry.

Evolution and Trends in Database MarketingThroughour this book, we have attempted to provide examples of marketingdatabases that reflect a wide range of industries and product categories. Withthe advent of new technologies and markets, the application of marketingdarabases will grow. Alrhough it is not feasible here to review all applicationsof marketing databases, we summarize below some of the key trends of themajor categories. Note that the categories are not mutually exclusive, that is,an organization in the b-ro-b or not-far-profit market can offer services.

Consumer Databases and the Internet

The growth of consumer databases for Inrerner marketing will besubstantial as more consumer purchases come from Internet sites. Analysistechniques will have to be developed to respond specifically to thisemerging marker. At least two categories of analyses are emerging, one foridentified customers and another for unidentified customers or visitors to

the Web site. Many lnternet databases have a unique characteristjc notfound in databases built for orher media: customers directly enter data intoInternet databases. The data entered sometimes include psychographic data~n. ~ddition to COntact information. The advantage to these custo01er-initiated databases is thar the databases may reflecr a stronger affiniry tothe brand or product category relative to darabases (lists) that are derivedfrom other indirect Sources.


Exhibit 17.4 Political Action Letter

February 7, 2000United States House of Representatives

Washington, DC 20515

Dear:

In the coming weeks, you may be asked to cast a vote on H.R. 354, the "Collections ofInformation Antiplracy Act." As the House of Representatives contemplates the issue of database

protection, the members of NetCoalition.com_an Internet industry trade group whose membersinclude Amazon.com, AOL, lnktomi, Yahoo!, Lycos, and Excite@Home_ask you to only support

legislation that balances reasonable protection with the equally important objective of allowing thefree flow of information and the legitimate access to information that are the cornerstones of the

Internet.We are concerned that H.R. 354 does not strike the appropriate balance and therefore poses

undue risks to the future growth of the Internet. NetCoaJition.com has endorsed an alternative,H.R. 1858, the "Consumer and Investor Access to Information Act of 1999," as introduced byRepresentatives Bliley, Dingell, Tauzin, Markey, Oxley and Towns and would prefer strongly that

Congress support this approach.NetCoalition.com includes some of the largest creators and publishers of databases. These com-panies recognize the concerns of database proprietors who object to theft and dissemination of

their databases.However, we also recognize that one of the Internet's greatest values lies in its ability to pro-

vide efficient access to information gathered from a multitude of disparate sources. Consumerscan use the Internet to find airline fares and schedules, music and movie reviews, sports starls-tics and results, stock quotes, research and analysis, and restaurant information. Overly broadprotection for compilations of information could lead to claims that Internet data collection is

unlawful, threatening the very Internet activity that consumers have found so beneficial.In considering this issue, it is worth noting that current law already provides effective means toprevent database theft. Copyright law provides remedies against those who engage in the copy-ing of databases. Contract law allows proprietors to impose limitations on the misuse of infor-mation. Technologies exist to prevent unlawful access to collections. Further, federal and state

laws prohibiting unauthorized access to servers linked to the Internet are already on the books.If Congress decides that additional database protection is needed, that remedy should be narrowlycrafted to address specific problems. We believe that H.R. 1858 meets this test, achieving the nec-essary balance between protection and availability of information. Unfortunately, other database

legislation pending in Congress will not achieve that critical balance.We appreciate your attention to this matter. Please let us know if we can prOVidefurther infor-

mation or answer any of your questions.

Sincerely,Daniel EbertExecutive Director

Source: Reprinted by permission of NetCoalition.


A challenge for many direct marketers is to integrate the database acrossmedia, so that transactions can be tracked regardless of where theyoriginate (e.g., mail, telephone, Internet, brick and mortar). The benefits ofintegrated databases to the marketer are substantial; one of them is beingable to track changes in how the consumer wishes to receive informationfrom the company. From the consumer's perspective, the integrateddatabase allows consumers to buy on the company's Web site and pick upor return merchandise at a local store.

Another trend in consumer databases is rhe promise of better dataanalysis. Advances in statistical methodology and computer technologyshould help marketers, using a broad range of direct marketing media, todevelop more efficient and effective marketing ptograms. Data andreports should be easier to access, and newer analysis techniques may beable to uncover and categorize meaningful complex behavior patterns onthe Internet.

B-to-B Databases

In the business market, marketing databases faU into the area of customerrelationship management (CRM). CRM software is a growing productcategory, and some estimates predict a continuation of a 50% annualgrowth tate (Trepper, 2000). Compared to consumer databases, b-to-bdatabases are usually smaller and require different analysis techniques.Some of the challenges for b-to-b databases are in the area of real-timeaccess to the database by multiple people within the organization. Forexample, an account manager on the road needs to have access to thedatabase to offer a customized proposal to a customer. At the same time, areorder of a standard component needs to be recorded in the database atthe home office by inside salespeople. Furthermore, with the enormousgrowth in b-to-b Internet transactions, an order or inquiry placed on theInternet needs to be recorded on the database.

Not-far-Profit Databases

In an increasingly competitive environrnenr, not-far-profit organizationswill become more dependent on good database development, mainte-nance, and analysis. Organizations such as hospitals will use databasesegmentation techniques to develop targeted relationship programs withgroups of patients and prospective patients. Not-for-profits that aretargetmg donors may have to develop new analysis techniques tomaximize communications with likely donors. Techniques that have


proven valuable for consumer producr marketing such as RFM (recency,frequency, and monetary value) may have to be modified for rhe nor-for-profit marker. For example, recenr major donors may feel rhar thecharitable organization is overly demanding or unappreciative if theyreceive numerous solicitations during the year. This approach mightnegarively affecr rhe long-term relationship with these donors.The challenge for not-for-profit organizations is to develop

relevant analysis techniques within tight budgetary restrictions. Because ofrhe unique relationships that are established berween charirableorganizations and its constituents, new models may need to be developedto maximize the effectiveness and efficiency of marketing programs. In thecase of most charitable organizations, the offer leaves the "price" open forthe donor to decide. Therefore, response to offers in yes or no terms maynot be rhe besr way to analyze the dara. Charitable organizations may needto use survey data to a greater degree in order to explore a variety of issuesor approaches that would maximize donation levels.

Retailer Databases

Retailers have been collecting an enormous amount of data for a numberof years. In particular, store cards used by supermarkets allow retailers totrack shopping patterns of individual custOmers. Translaring rhe data intomeaningful information for marketing programs has been a challenge forretailers. Many retailers don't have rhe capability to successfully mine thescanner data, but a few applications have emerged. For example, a SouthCarolina supermarket chain is using the customer database to reward cus-

, be: d t kiosk checktorners for past purchases. Customers swipe t err car a ~ I .'

their balance and receive certificates for free products. A WlsconSI11super-market chain uses transaction data from store cards to personalize shop-ping lists mailed ro members of a loyalty program (Lach, 1999),

Service Organization Databases, ' b d Most new businesses

The U.S. economy is beconung more service ase . .. .' s must be responSive to

are services. Because many service organlzauon .., d b h become a entlcal elementindividual needs of customers, ata ases ave . .for developing customer relationships. Databases for serviCe o.rgamza-, 'f P eferences and reqUJrementsnons may include a unique set 0 customer r ..(e.g. room and service preference for a hotel chain, style char~cte~lsr,c)S

, , 'II s and medicatIOns.for a home designer's customers, pauent I nesse


Databases with these types of variables that are more qualitative thanquantitative may require different analysis techniques. Software isavailable to examine the unstructured data (e.g., open-ended questionsabout service needs) that might be collected from customers seekingcustomized services.

Chapter SummaryDatabase marketers have to be attentive to a variety of issues in themarketing environment. Few organizations can ignore the globalization ofmarkets, but the challenges of global database marketing are substantial.The organization needs to be aware of all the implications of usingdatabases for marketing purposes in other countries. In particular,marketers have to be responsive to the legal and cultural environments inthe target country.

In recent years, database marketers have become the target of publicinterest groups, legislators, and litigation. Within this environment, it isnot sufficient for marketers just to adhere to the letter of the law.They have to examine how public interest groups and legislators aregoing to respond to their actions. Industry organizations such as theDMA have established guidelines in a number of areas that havean impact on consumers. The goal of industry organizations is to main-tain a code of conduct through self-regulation rather than throughlegislation.

In the final section of this chapter, we examine some of the trends indatabase marketing. With advances in technology, more organizations willbe collecting customer data. The growth of commerce on the Internet willfuel the development of many of these databases. In addition, because ofthe need for organizations, including not-far-profits, to become moreeffective and efficient, marketing database development and analysis willincreasingly be a mechanism to achieve that goal. The challenge fororganizations is to transform the data into information that assists businessdecision making.

Review Questions

1. Why are organizations moving into global markets?

2. Discuss the variables that database marketers have to consider inthe global environment.

3. With regard to ethics and public perceptions, what are the internaland external influencing groups?


4. What is an industry organization, and what are their primary goals?

5. What are some of the current public policy concerns of databasemarketing organizations?

6. Discuss some of the future trends rhat will affect database marketing.

Glossary

address standardization The process by which direct marketers clean upthe addresses contained on their database. This process uses CASS-cerri-fied software, which is required ro qualify for cerra in postal rate

discounts.aggregate marketing A merhod ro reach customers through mass mediaand traditional retail distribution channels that does not depend on data~rom individual customers bur rather aggregated level statistics andinformation.

analysis sample/data set A portion of the sample of customers on adatabase upon which the analysis for determining the target market isperformed. Typically, the portion of the sample upon which rhe analysiswill be conducted represents rwo thirds of rhe total sample. The otherone third is held out for validation (see validation sample).

attributes Customer information relating to purchase, promotion,demographics, and psychographies. Attributes are also known as data

fields or variables.banner ad Online advertising that usually appears as graphic images.When banner ads are clicked on, the user will often be taken ro another

Web location.batch processing A method of data processing usually used by organiza-tions with large databases on tasks rhat require substantial computermemory and processing capabilities. Batch processing occurs offline, andtherefore the changes are not made immediately ro the database butrather are made ar periodic set schedules (e.g., biweekly, weekly,

monthly).binary data Data that are discrete and only take on two values. Forexample, the variable "own home" can either take on rhe value of yes or

no.block groups Subdivisions of census tracts formed by grouping blocks(streets). There are approximately 225,000 block groups.


bootstrapping A method of building a stable regression model when ade-quate sample sizes are not available. This technique is also referred to asbagging.

breakeven response rate The response fate required in which allpromotional costs are covered and no profit is realized.

brick and mortar retailing Retailers who have physical locarions.cannibalism A situation that occurs when sales of a company's currentproduct base are taken away by the introduction of new products.

CASS U.S. Postal Service Coding Accuracy Support System-certifiedsoftware used in the process of standardizing addresses on a database.

categorical variables Data that take on a finite (countable) number ofvalues and are descriptive in nature with no meaning relative to oneanother. For example, "car type" is categorical and may take on valuessuch as 1 ~ Sedan, 2 = SUV,3 = Compact.

census data Data gathered by the government every 10 years that includesaverage or median income levels, home values, and so all, within each zipcode, block group, and/or census tract.

census tracts Subdivisions of counties as defined by the U.S. government.Today there are approximately 50,000 census tracts.

CHAID (Chi-squared Automatic Interaction Detection) A statisticalmethod of determining statistically meaningful splits in various datafields or variables. It is also sometimes ref erred to as a tree algorithm.

clicks and bricks Organizations that have traditional physical outlets(retail stores) and are also involved in Internet commerce.

clicks and mortar Sec clicks and bricks.clickstream The sequence of clicks or pages requested as a visitor exploresa Web site.

clickstream data The data collected by organizations to determine thesequences or paths that visitors take as they explore a Web site.

clone models A regression model that is built for the specific purpose ofidentifying prospects on an outside list who look like your bestcustomers. Such models are also called best customer models or matchmodels.

coefficient of determination With respect to regression modeling, it is themeasure between 0 and +1 that tells you the percentage of variation inthe response variable explained by the introduction of the predictorvariables.

coefficients The weights associated with each of the predictor variables ina regression model.

compiled list data Lists of people or organizations gathered fromtelephone directories, voter registration files, membership rosters,department of motor vehicles, surveys, and so on.

computer network A system of connecting computers and peripheraldevices together with SOftware and hardware. Networks can help

Glossary 371

improve the effectiveness and efficiency of marketing activities ballowing quick access to relevant data. y

confidence interval calculation A statistical technique that reveals basedon a test, a range in which the true response rate in rollout is likelyto fall.

contmuous vaflab.les Variables that can rake on any value within a rangeof values. TYPIcally, the values of the variable have some meaningrelative to one another. Examples include age, ratios, product paidcounters, average cellular minutes per month.

cookie Computer code that is placed on the visitor's hard drive by a Website so that each time the visitor returns to that Web site, they can berecognized and tracked.

correlation analysis A statistical procedure used to determine the strengthof association between any rwo variables.

correlation coefficient A value between -1 and +1 that reflects thestrength and direction of a relationship between two variables.

cross-tabulation A tabulation displaying two or more data elements incombination, highlighting interrelationships among variables.

Customer contact data The data necessary for a company to reach acustomer (e.g., name, address, phone and fax numbers, e-mail address).

Customer-initiated database A database in which the customer hasinitiated the relationship by contacting the marketer, rather than themarketer attempting to initiate the relationship through mail ortelephone contact.

database A collection of information related to a particular subject orpurpose that is usually maintained on a computer for easy search,retrieval, and analysis.

database integration Systems for synchronizing and coordinatingdatabases across different media (e.g., mail, Internet, retail).

database management system Software and hardware that allowinformation to be created, modified, and accessed more efficiently.

database marketing Marketing activities (e.g., selecting prospectivecustomers) that use a marketing database.

data enhancement A process by which external data are "overlaid" orappended to an existing database such as a house file for the purpose ofunderstanding more about customers and increasing the effectiveness of

marketing programs.data fields See attributes. .data mart A repository of data designed to serve a particular commullityof knowledge workers. In scope, the data may derive [rom an

d h 0 be more specialized. Theenterprisewide database or ata ware ouse r . . aemphasis of a data mart is on meeting the specihc de~lands ofparticular group of knowledge users in terms of analysis, content,

f U f d ta mart can expect to havepresentation, and ease a use. sers 0 a adata presented in terms that are familiar.


data mining The process of identifying previously unknown relationshipsand patterns in data, in part.icular customer databases, in order to solvea business problem.

decoy records Owners of databases often include names in the file thatare not real customers (decoys) to check how the mailing files are beinghandled by outside service bureaus or renters of a list.

deduping the customer file The process by which direct marketersidentify duplicate customer records and combine and eliminate themfrom the database or mailing file.

demographics Characteristics such as gender, age, income, familycharacteristics, and occupation that are used to describe and segmentcustomers.

dependent variable In regression modeling, this is what we are trying topredict (e.g., response, payment). This variable is also called the responsevariable.

direct marketing An interactive system of marketing that uses one ormore advertising media to effect a measurable response andlortransaction at any location, with this activity stored on the database.

Direct Marketing Association (DMA) A professional organization thatserves the direct marketing industry by providing educational programs,research about the industry, privacy and other guidelines, politicalsupport for critical issues. They can be found at www.the-drna.org.

discount rate The net present value (NPV) of an investment adjustsreturns on that investment by a certain percentage rate (discount rate) toreflect its true value in today's money.

discrete variables Similar to categorical variables in that they take ononly a very finite and countable number of values, but in this case thevalues themselves have some meaning relative to one another. Examplesinclude number of children and number of cars owned.

diversification strategy The process of moving into new markets withnew products. Testing the new products on a selected database ofpotential Customers may be a way to reduce diversification risk.

DMA Preference Services A voluntary program used by many directmarketers to purge their files of individuals who requested that their namesbe removed from direct marketing databases. There are preference servicesfor mail, e-mail, and telephone databases.

dry testing Making a mock offer to determine the extent of the demandfor a product prior to finalizing product development. There are legalconstraints on how the mock offers can be worded to make consumersaware that the product may not be available.

e-commerce (electronic commerce) Business activity (e.g., buying, selling,communicating, servicing) that uses electronic media such as theInremet.

e-mail (electronic mail) Used for communicating on the Internet.

Glossary 373

E-Mail Preference Service (E-MPS) A list of names maintained by theDMA of people who prefer not to receive promotional offers via e-mail.

encryprion Data-coding techniques used to protect files and individualrecords from unauthorized use. Almost all e-comrnerce companies useencryption as a means of protecting sensitive customer data.

enhancement data Customer information gathered by outside sourcessuch as compiled, response, and modeled data, Also known as externaldata.

expected profit A value derived by combining one's likelihood of orderingand paying with the dollar values associated with such actions,

external data See enhancement data,factor analysis A statistical analysis method that determines the factors or

constructs that underlie the data and reveals relationships that are not eas-ily observable. It is also know as principal components analysis or PCA,

factors The unique groups of customer data elements determined byfaeror analysis based on patterns observed,

frequency data Customer data related to a customer's total number ofpromotions, orders, payments, and so on.

frozen file A sample used to determine the unique characteristics thatdistinguish responders from non responders, On such a file, the charac-teristics of each customer is reflective of how they looked at rhe time thepromotion was sent.

fulfillment data Data relating to the fulfillment of a customer order orother activity (e.g. "date of product shipment" is a fulfillment dataelement).

fulfillment file The file containing all customer fulfillment data, All directmarketers must have a fulfillment file or database to conduct business.

fuJi factorial test design When conducting marketing tests with severalpackage elements, sometimes a direct marketer is interested lJ1 assessl~gevery possible package configuration, Doing so is called a full factorialtest design. f

gains chart A table or graph showing the expected response of groups 0CUStomers to an offer based on a predictive model. [ndividuai customersare scored according to the predictive model, and the file is sorted, usu-ally into 10 groups or deciles, each representing 10% of the total sam~le.

geo-demographic data A c1assificarion of customers based on where t eyI· 'b di rive of purchase patterns.ive. Where a customer lives may e pre rc I . k

global business environmeot The process of evaluating potential mar ets

in other countries, j d b Hardwareh ' h holds tne ata ase.ardware The physical eqUipment t at devi d

devi . ut and output evices, anincludes processors, storage evrces, uipcomponents that link devices together in networks, d d ( 'thin

hot-line lists Lists of customers who have recently respon e e.g., WI

the last 90 days) to some type of offer,


house data Internal or house data is data obtained from sources withinthe organization. For example, customer contact data) past purchaserecords, product returns data, and customer services data all constitutehouse data.

house file A database of existing customers that usually containsextensive house or internal data regarding responses to past marketingprograms. External demographic and psychographic data may also beincluded in the file.

householding the customer file The process by which a direct marketeridentifies individuals on the customer database residing at the sameaddress for purposes of obtaining promotional efficiencies.

hypothesis test A statistical technique used to determine if a particulartest response rate has beaten the control response rate.

independent variables In regression modeling, these are what we call thevariables or data dements that arc being used to predict response,payment, and so on. Such variables are also called predictor variables.

interactive marketing Often used interchangeably with direct marketing.internal data Customer data collected internally on customers, includ-

ing all contact, promotion, order, and monetary data. Also see housedata.

Internet service provider (ISP) An organization that provides customersaccess to the Internet.

key codes Codes that indicate which individuals received particularmarketing programs. Key codes allow marketers to evaluate the per-formance of each marketing program.

legacy systems Database systems that are many years old, often spanningdecades. These systems are updated on a regular basis, bur they are notstate-of-the-art database systems. To convert the entire system over to astate-of-the-art system would be either too expensive or too disruptivefor the business.

lifetime value (LTV) The value today of future profits from customers.LTV is used to determine the return to the company for making an invest-ment in gaining new customers, specific marketing programs, and prod-uct lines.

lifetime value (LTV) analysis The methods used to determine the return tothe company for making an investment in gaining new customers,specific marketing programs, and product lines.

logic counter variables A variable that represents how many' of a certaincriteria are met by each customer. For example, a count could be createdrepresenting the total number of music genres each customer checked ona survey.

logistic regression A different form of regression modeling that yields atrue probability of the Customer taking a specific action (e.g., respond,pay, renew).

Glossary

longitudinal or time series variables Variables that allow direct marketersto view a particular data element for each customer across time. Forexample, one way to estimate a customer's action on the next promotionsent may be to examine the customer's response (order, pay, silent, etc.)to the last three promotions sent to them.

LTV See lifetime value.Mail Preference Service (MPS) A list of names maintained by the DMA of

people who prefer not to receive promotional material by mail.manual selects A simple definition of the target market typically based on

two or three key customer characteristics. For example, selectingcustomers from the database that are single, female, and age 30-50would be considered a manual select.

market development strategy The offering of existing products to newmarkets. A database of customers in other markets could be acquired torest the potential of a product in those markets.

marketing data Any piece of customer information such as past purchaseinformation used by marketers for the purpose of increasing theeffectiveness or efficiencies of marketing activities.

marketing database A file or group of files containing information aboutcustomers that enhances the market.ing process. With current technology,the database is stored, manipulated, and analyzed on a computer.

marketing objectives Quantitative and time-specified targets for businessand marketing activities. Objectives may be specified for sales volume,profitability, market share, brand awareness, intensity of distribution,and so on.

marketing planning A process that includes performing a situationalanalysis of markets, specifying marketing objectives, developingmarketing strategies, implementing tactics (marketing programs), andmonitoring and controlling the process.

marketing strategy The longer-term direction to influence customers andachieve marketing objectives. It involves developing products to meetCustomer needs and positioning products (i.e., communicating aboutbenefits) to target segments.

marketing tactics The specific actions (or programs) to implement the mar-keting strategy. Each element of the marketing mix (promotion, pric~, dis-tribution, product) should be considered in rhe development of tactics.

market penetration strategy Increasing product use for exrsnng custom.ersor noncustorners with similar profiles. A database may assist in reachingexisting customers in a more efficient and effective manner.

match code A number and letter code that uses elements of a name andaddress to develop a unique identifier of an individual or household on a

database.match coding The process by which match codes are assigned to each cus-

tomer record for purposes of deduping or householdmg the customer file.

375


m-commerce (mobile commerce) The marketing of goods and servicesthrough wireless handheld devices such as cellular telephones andpersonal digital assistants (pDAs).

merge/purge processing The process of deduping mailing lists associatedwith an outside list campaign and purging the duplicate records.

modeled data Data generated from statistical analysis such as customerclustering according to demographic, psychographic, or past purchase data.

monetary data Customer data related to a customer's total dollar value oforders placed, payments made, and so on.

monitor and control Comparing actual performance with marketingobjectives and making adjustments to aspects of the marketing plan asneeded. If performance does not reach object.ives at specific milestones,the program or strategy may be modified.

multicollinearity Problem that OCCUtS when building a multipleregression model in which strong correlations between the predictorvariables exist. Such a problem can lead to an unstable model.

multiple logistic regression A different form of regression modeling thatyields a true probability of the Customer taking a specific action (e.g.,respond, pay, renew).

multiple regression A statistical method that builds a predictive equationor model based on the "best fit" between a dependent and one or moreindependent variables.

mu.ltivariate analysis A statistical technique that reveals unusual and notreadily apparent relationships in the customer data. Types include factoranalysis, cluster analysis, and discriminate analysis.

NCOA (National Change of Address) A U.S. Postal Service-approvedservice that updates address changes and removes nondeliverable mailfrom direct marketing files. NCOA maintenance must be conducted forcertain special postal discounts.

negative correlation A relationship in which higher values of one variableare associated with lower values of another variable.

net present value (NpV) The net present value of an investment adjustsreturns on that investment by a certain percentage rate (discount rate) toreflect its true value in today's money.

neural networks A statistical method that "learns" the data byexamining patterns. This technique is most common in credit frauddetection.

nixies Nondeliverablemail that should be removed from a database toeliminate unneeded expense.

nth selects A method by which marketers sample the database to ensurerandom selection of Customers. For example, to test a new format to5,000 names from a database of size 100 000 the direct marketer beginsby selecting 1 name on the database, choosing every 20th(100,000/5,000) name thereafter.

Glossary377

opt-in ,A ~olicy requiring that individuals give permission before anorga.n1zatlOn ,sends o,Hers or information through a direct marketingmedium (mad, e-mail, phone). With this policy no solicitations orcomn:unications will come to individuals without their expressedperrrussion.

opt-out A policy in which individuals have to request that no oilers orinformation be sent to them through a direct marketing medium (maile-mail, phone). With this policy, solicitations or communications will besent to an individual unless they take a specilic action (e.g., filing a form).

outsourcing Using external organizations or individuals for variousservices such as database maintenance, analysis, mailing, and orderprocessing.

point-in-time data A sample comprised of point-in-time data is used todetermine the unique characteristics that distinguish responders fromnonresponders. On such a file, the characteristics of each customer reflectshow they looked at the time the promotion was sent. (See frozen file.)

positioning The process of establishing and maintaining a certain imageof a company's product, relative to competitors, in the customer's mind.

positive correlation A relationship in which higher values of one variableare associated with higher values 01 another variable.

predictor variables In regression modeling, these are what we call thevariables or data elements that are used to predict response, payment,and so on. They are also called independent variables.

product development strategy The process 01 developing new productslor existing customers. The database can help in product developmentthrough the use of systematic testing paradigms.

prospecting database A database comprised solely of noncustorners.Typically, names and addresses found on such files come from compiledlists.

psychographies Activities, interests, and opinions of individuals such ashobbies, recreational activities, and political and social opinions, and soon that are used to describe and segment customers.

purchase behavior These data include previous purchases by productcategory, payment history, purchase frequency, and amount. Thisinformation is often predictive of future purchases.

pure play e-commerce site A company that only uses e-commerce asopposed to multichannel marketers.

purge The removal 01 certain records or data items from a database orlist.

random sample A sample in which every member of the sample is equallylikely to have been chosen, ensu.ring a composinon similar to that of rhepopulation from which the sample was drawn.

ratio variables A variable derived by dividing one data e1enl,ent b

hy

. f rh average payment rate or eacanother. For example, an esnrnate a e


customer could be derived by dividing total products paid by each cus-tomer by total products otdered.

real-time processing A method of data processing in which task requestsare entered into the system immediately and processed according to useror task priorities.

recency data Customer data related to the receocy of a customer's lastpromotion, order payment, and so all.

reconciliation The process of comparing forecasted to actual campaignstatistics such as response rates, payment rates, and profit.

regression analysis A statistical method that builds a predictive equationbased on the "best fit" between a dependent and one or more independ-ent variables. It is also known as regression modeling or responsemodeling.

regression modeling See regression analysis.relational databases Databases that have no predetermined relation-ships between data items. Information is contained in tables that caninteract with each other. Relational databases are more flexible for dataaccess.

representative sample A sample accurately reflecting the population ofinterest from which direct marketers draw inferences. For a sample to berepresentative, no members of the population of interest are purposelyexcluded.

response lists Lists of people or organizations that responded to offerssuch as mail order catalogs, subscriptions, or solicitations for donations.

response modeling Methods (e.g., mathemarical or statistical) used toselect customers most likely to exhibit a particular characteristic (e.g.,ordering, paying, renewing). It is also known as predictive modeling.

response variable See dependent variable.reverse test When rolling out with a new format or creative concept,direct marketers often retest (reverse test) the old promotional package.Doing so provides valuable information regarding the performance of thenew test package in rollout.

RFM (recency, frequency, and monetary) data elements These variablesare used to predict future purchases.

RFM analysis/scoring A method of assigning certain values and weightsto past purchase activities (recency of purchase, frequency of purchase,and the monetary amount of purchase) in an attempt to rank customersfrom those most likely to order to those least likely to order. One canchoose from several scoring algorithms.

rollout A large-scale direct marketing campaignsalting Owners of databases often include names in the file that are notreal customers (decoys) to check how the database is being used byoutside service bureaus or renters of a list. This process is known assalting the file.

Glossary 379

sample A sampl~ is a subset of customer records. Inmost cases, it is takenrandomly and IS representative of the universe of interest on directmarketers' databases.

sample size estimation When planning a test of marketing programs,analysts should calculate the required sample sizes for the tests to ensurea certain level of accuracy in the results.

SAS Statistical analysis software commonly used by database analystsdeveloped by the SAS Institute.

scoring The process of attaching a value onto a customer record based ona predictive model such as regression at a formal RFM analysis.

segmentation A process of dividing a market into smaller pieces based ondemographic, psychographic, or behavioral (purchase) patterns.Segmentation is necessary to develop marketing plans that are moreresponsive to a specific group of customers.

SIC (Standard Industrial Classification) codes Four-digit numerical codesassigned by the U.S. government to business establishments to identifythe primary business of the establishment. The classification wasdeveloped to facilitate the collection, presentation, and analysis of dataand to promote uniformity and comparability in the presentation ofstatistical data collected by various agencies of the federal government,state agencies, and private organizarions.

simple linear regression A regression modeling method that uses oneindependent variable and one dependent variable.

situational analysis An evaluation of environmental factors (cultural,economic, legal, political, social, demographic, technological, and soon), present and future markets, target market characteristics, competi-

tors, and so on.snapshot A file used to determine the unique characteristics that

distinguish responders from nonresponders in which the characteristicsof a II the customers on the file reflect how they looked at the rime the

promotion was sent. (See frozen file.)Source A code that indicates how an individual record (i.e., name, house-hold) first entered the database. Sources include lists purchased by themarketer, response to advertisements, cards filled out at rerai.l,and so on.

spam Unsolicited or unwanted commercial e-mail. Spammlllg refers tothe process of sending unsolicited commercial e-mail.

SPSS Statistical analysis software commonly used by darabase analysts

developed by SPSS Inc. I . Istepwise regression A particular method of building any mu np e

. . I . hi' bl t meet a specific thresholdregression model IJ1 w lIC eac 1 varra emusprior ro entry into the final model. . .

structured database A database that has defined relationshiPS and paths.Data in structured databases come from a single source. lrerns on thedatabase are linked one ro one: only one item relates to another.


structured query language (SQL) A computer language used for databasemanagement.

summary/aggregate data Data that provide marketers with informationon how marketing programs are performing by totaling or averagingdata from individual records. Tara! orders placed in response to a specificmailing is an example of aggregate data.

target market The exact definition of a customer group deemed mostappropriate for a particular product/service or promotional offer athand. Database marketers often define target markets via analysis of pastpromotional tests.

Telephone Preference Service (TPS) A list of names maintained by theDMA of people who prefer not to receive promotional offers via thetelephone.

transaction data Variables that relate to product purchase or otherresponses such as buyers versus non buyers, recency and frequency ofpurchase, amount of purchases, brand loyalty, position in adoption cycle,and product attitudes.

univariate tabulations Tabulations produced on analysis samples and dis-play tbe percentage responders and non responders to the offer for thevarious categories of each data element.

validation sample/data set A portion of the sample of customers uponwhich the analysis is validated prior to using such findings for selectingnames for promotion. It is also called the hold-out sample. Typically, theportion of the sample upon which one validates the analysis findingsrepresents one third of the total sample.

variables See attributes.

zero correlation A relationship in which higher values of one variable areassociated with all values of another variable and vice versa. Zero or nearzero correlations indicate that two variables are not related to each other.

Additional Readingsin Databaseand D ired Marketi ng

Database and Direct Marketing

Beyond 2000- The Future of Direct Marketing, by Jerry ReitmanThe Complete Database Marketer: Second-Generation Strategiesand Techniques

for Tapping the Power of YOlfr Customer Database, by Arthur M. Hughes.Customer-Driven. Marketil1g, by John Frazer-RobinsonCustomer Relationship Management: A Senior Management Guide to Technology

for Creating a Customer-Centric Business, conducted by Price WaterhouseCoopers, commissioned by the Direct Marketing Association, Inc.

Database Marketing: The New Profit Frontier, by Ed BurnettDatabase Marketing: The Ultimate Marketing Tool, by Edward NashData Minil1g Your Website, by Jesus MellaThe Data Warehouse Lifecycle Toolkit: Expert

Developing, and Deploying Data Warehol/ses,Reeves, Margy Ross, & Warren Tbornthwaite

Desktop Database Marketing, by Jack Schmid & Alan WeberDirect and Database Marketing, by Graeme McCorkellDirect Marketing: An Integrated Approach, by William J. McDonaldDirect Marketing Managemen.t, by Mary Lou Roberts & Paul D. BergerDirect Marketing: Strategy. Planning, Execution, by Edward NashThe Engaged Customer: The New Rules of Internet Direct Marketillg, by Hans

Peter Brondmo & Geoffrey MooreHow to Find and Cultivate Customers Through Direct Marketing, by Martin Baier

Integrated Direct Marketing, by Ernan Romand S

. ( C t ter RelationshipMastering Data Mining: The Art an, ctence 0 us on

Management, by Michael J. A. Berry & Gordon Linoff

Methodsby Ralph

for Designing,Kimball, Laura


The New Direct Marketing: How to Implement a Profit-Driven Database MarketingStrategy, by David Shepard Associates, Rajeev Batra (Ed.)

The New Integrated Direct Marketing, by Mike BerryThe Next Step in Database Marketing-Consumer Guided Marketing, by Dick

Shaver

Power of Your Customer Database, by Arthur M. HughesStrategic Database Marketing, by Robert R. Jackson & Paul WangStrategic Database Marketing: The Masterplou for Starting and Managing a

Profitable Customer-Based Marketing Program, by Arthur M. HughesSuccessful Direct Marketing Methods, by Bob Stone

Statistics ReferencesAIJplied Linear Statistical Models, by John Neret; Michael Kutner; Christopher

Nachrsheirn, & William Wasserman

Applied Regression Analysis and Other Multivariable Methods, by David Klein-baum, Lawrence Kupper, & Keith Muller

Data Mining Cookbook, by Olivia Parr RudA Second Course in Statistics: Regression Analysis, by William Mendenhall & Terry

Sincich

References

A better supermodel than Kate Moss. (1999, November 15). Direct.Allstate unit establishes information technology center in Northern Ireland; notes

lack of U.S. computer programmers. (1998, October 24). Insurance Advocate.

109(42),26.Ansoff, I. (1988) The new corporate strategy. New York: Wiley.Baier, M" & Stone, B. (1996). How to cultivate customers through direct market-

ing. Lincolnwood, IL: McGraw Hill-NTC.Beardi, C. (2001, April 16). CRM. Advertising Age.Belfer, S. (1998) IT are from Mars, and marketing are from Venus. Direct

Marketing, 61(5), 52.Blank, C. (2001, March 5). Studies say l11-CQmmerce market is Ullpredictable.

eMarketer.Book, J. (2000, June 2). A multichannel effort yields the best results. iMarketillg

News, 25,Brenner, K. (2001, June 25). Report: Privacy inadequate on government Web sites.

iMarketing News. 3.Brenner, K. (2001, July 9). Poll: Privacy not top social issue for consumers.

iMarketing Neios, 4.Campanelli, M. (2001. June 25). Joint congressional panel to weigh difference in

privacy legislation. DM News. 3.Cluster analysis helps Proflowers personalize email, increase profits. (2001,

May/June). ltol Magazine.Cruz. W. (2001, March 21). Kawasaki gets big response from rich media e-mails.

iMarketing News.Customer service (2001). Retrieved from www.payless.com

CYS launches first integrated promotion. (2000, ApriI1l). dmnews.com.

Dana, J. (1999, September 13). HIT the bricks. Marketing News, 33(19),1.Dana, J. (2000, July 17). New technologies in marketing. Marketing News, 25-28.Davis, J. (2001). This year may become the year that bricks-and-clicks achieve their

revenge. In(o World. 23(3),70.


Dell corporate information (2001). Retrieved from www.dell.com/us/en/gen/corporarelfacrpack_003.hrm

Direct Marketing Association (2000). Economic impact: U.S. direct marketing today(6th ed.) (chap. 1). Retrieved from www.the-dma.org/library/publicationsllibres ecoimp l.b Ia.shtml

Drozdenko, R.) & Cronin. J. (2001) [Consumer perceptions of mulri-channel pur-chasing options]. Unpublished.

eMarketer. (2001). Retrieved from www.ernarkerer.comFingerhut to launch new e-mail, database marketing program. (2000, April 11).

dmnetus.com,First Union banks on fibre optic expertise from Amdahl. (2001). Retrieved October 1,

2001 from www.amdahl.com/sllccess/mm002891.hrml

Full-sizesport utility market import vs. domestic = status vs. utility. (1997,january 6).PR Newswire.

GoldMine. (2001). Retrieved from goldmine.com/supporr/care/isolutions/ case-study.cfm?caseid = 22

Gordon, C. (2001, June in. Establishing a Web presence in the UK. DM News, 32.Haegele, K. (2000, March). Hispanic Americans, a crash course in culturally sensi-

tive marketing. Target Marketing. 23(3), 97.IBM (2001). IBM zSeries (formerly 5/390) solutions for business intelligence. Retrieved

from www-4.ibm.comfsoftware/dara/bils390/solutionslindex.htmJackson, R., & Wang, P. (1997) Strategic database marketing. Lincolnwood, IL:

McGraw Hill-NTC.

Krol, C. (1998). Weber segments customers to keep its grill sales sizzling.Advertising Age. 69(36), 17.

Lach, J. (1999). Data mining digs in. American Demographics. 21(7), 38.Levey,R. H. (2000, August). Pleasantville, NY: Pop. 30 million. Direct.Levey, R. H. (2001, July). Just for you. DirectLivePerson and Harris Interactive unveil the post-holiday customer satisfaction

report. (2000, February 10). Business Wire.Long, C. (2000, May). You don't have a strategic plan?-Good! Consulting to

Management C2M, 11(1), 35-42.Mena, J. {I999). Data mining YOllrwebsite. Boston: Digital Press.Mistry, S. (1996, April). To succeed in India, marketers must look beyond the num-

bers. Advertising Age International, 16.Molloy, C. (1997, November 3). Lloyds, Ciribank rake dead aim on Brazil. Global

Fund News. 1(11), 1-2.MotherNature.com streamlines its e-mail efforts. (2000, May 19).dmnews.com.Orr, A. (2000, February). Count me in-apt-in, that is. Target Marketing. 23(2), 5.P & G makes AOL debut with mouthwash ads. (1999, March 1). Marketing

News, 1.

Pastore, M. (2001, June 18). Women maintain lead in Internet use. CyberAtlas.Porter, M. (1001, March). Strategy and the Internet. Harvard Business Review,

63-78.

References 385

Priore, T. (2000, September 7). Improving e-mail response rates. DM News.Regression modeling turns Carcer'Track's DM campaign from static to dynamic.

(1998, March 12). drnnews.com.Roberts, M., & Berger; P. (1999). Direct marketing management (2nd ed.). Upper

Saddle River, NJ: Prentice Hall.Saturn Corporation. (2001). Retrieved from www.saturn.comSchmit, 1- (1999, February 16). Asia's culture hampers Internet commerce. USA

Today, p. 6B.Schulrz, R. (2001, May 15).CRM cynics. Direct.Shermach, K. (1998, June). Zen and the art of marketing cards to Asian consumers.

Card Marketing. 2(6), 16-17.Stone, B. (1996). Successful direct marketing methods (erb ed.]. Lincolnwood, IL:

McGraw Hill-NTC.Surowiecki, J. (1999, February 1). The return of Michael Porter. Fortune, 139(2),

135-137.Sutter, M., & Mandel-Campbell, A. (1998, Ocrober 5). Customers are eager, infra-

structure lags. Advertising Age tnternational, 12.Tiffany & Company strikes gold with AS/400 business intelligence solution (2001).

Retrieved from www_l.ibm.com/servers/eserverliseries/casest/tiff2.htmTomasula, 0.(2001, May 10). Study: $2.:1 billion to be spent on e-mail marketing

in 2001. iMarketing News.Trepper, C. (2000, May 15). Customer care goes end-to-end. In(ormatiollWeek,

786,55-61.Weaver, J. (2000, August). New economy lies. Smart Businessj 103-8.Wunderman, L. (1998, October). Keynote address to Direct Marketing Association

of Washington.Yorgey, L. (1998). Global media choices. Target Marketing, 21(4), 29.You say London, I say Londres- (1997, May). Marketing Toofs, 12.

Name Index

Ansoff, I., 30 Mistry, S., 355Myers, J. H., 163

Naik, D. N., 163Baier, M" 146Beardi, c., 7Belfer, S., 86Berger, P. D., 36Blank, c., 320Book, J., 310Brenner, K., 319,360

Campanelli, M., 360Cronin, J., 319Crozdenko, R" 319Cruz, W, 322

Orr, A., 354

Pastore, M., 313Polk Company,S, 6, 203Porter, M" 26Priore, T., 322

Roberts, M. L., 36

Schmit, J., 353Schultz, R., 7Sherrnach, K'J 353Stone, B., 34Surowiecki, J., 26Sutter,M., 353, 357

Dana, J., 44, 320Davis, J., 310eMarketer, 312, 313, 320, 321

friedman, J. H., 226 Tornasula, D., 321Trepper, C, 364Gordon, C, 354

Haeglel, K., 352Hall, 1'., 226Harris Interactive, 314Hughes, A., 145, 146

Jackson, R., 85, 92

Wang, P., 85, 92Weaver,].,318Wunderman, L., 8

Yargey, L., 357

Khattree, R., 163Krol, c., 132Lach, J., 365Levey, R. H., 5, 51Long,C.,2S

Mandcl-Campbel, A., 353, 357Mena, J., 335

3 7

Subject Index

Access) 12Account management, ] 0-] 1ACT! 12,81,82 (exhibit)Acxiom,50Aggregate marketing, 14, 16 (exhibit)Airline Hydraulics, 133Allstate Insurance 356America Online, 344Analysis samples, 95~96

applications, 101creation of, 98-99frozen files in, 98point-in-time customer data, 98-100,

100 (exhibit)random samples, 97representative samples, 96-97, 96

(exhibit)utility of, 97-98validation samples, 101

Analysis. See Marketing test results;Statistical analysis; Strategicreport-ing/analysis

Ashford.com, 343

Banner ads, 134, 318, 345-347, 346(exhibits)

Breakeven response rate, 108, 109increased rate, 283-284required rate, 283

Brick-and-mortar businesses. See Retailtransactions

Business market organization, 11, 364

Cannibalism, 164-165Career Track, 184CASS-<:ertified software, 61Census data, 47-49, 48 (exhibit)Center for Science in the Public Interest, 12CHAID (chi-squared aurornated

interaction detection) analysis,148-153,149-152 (exhibits)

C?ildren's Online Privacy Act, 319Cisco Systems, 315-316Clicks and mortar retailers, 310Coding. See Database maintenanceCommunications. See Marketing

communicationsComputer technology, 11-12

See also Database technologyConfidence interval calculations, 252

business decision-making and,266-268

confidence level,determination of,263-265

sample mean formula, 253~256, 255(exhibits)

sample proportion formula, 256-258two sample means, difference

between, 258-260, 259 (exhibit)two sample proportions, difference

between, 260-263, 262 (exhibit)Cookies, 320, 330-331Correlation analysis, 120-123, 121

(exhibit)correlation coefficient, 126-127, 126

(exhibit)negative correlation, 123-125, 124

(exhibits)positive correlation, 123, 123-]24

(exhibits)zero correlation, 123, 125 (exhibits)

The Credit Index, 67Credit risk/fraud, 67, 206CRM,6-7Cusromer-centric marketing, 6-7customer-initiated databases, 362, 364customer sophistication, ]2-13respollsive policies/practices, 35-36See also Online customers

Customer data, 40activity data, 43-44, 98, 130census data, 47-49, 48(exhibit)


compiled lise data, 45-47, 46 (exhibit),49-51

customer contact data, 44-45database typology, 41data needs determination, 40external/enhancement data, 45-49, 50,

51-53fulfillment data, 41, 43internal/house data, 41-42, 42

(exhibit)marketing data, 43-44, 98modeled data, 49point-in-time data, 98-100, 100

(exhibit)prospecting data, 24, 41See also Database maintenance; Online

CustomersCustomer data analysis, 103-104

breakeven criteria, 108, 109correlation analysis, 120-1.27, 121

(exhibit), 123-126 (exhibits)cross-tabulations, 111-113, 112

(exhibit)familiarization process, 104-105key events, time alignment of, 119-120logic COunter variables, 113-119new-co-file customers, 109promotional counter variables, 111promotional intensity and, 163-164univariate tabulations, 106-111, 107

(exhibits), 1.09-110 (exhibits)See also Market segmentation;

Strategic reporting/analysisCVSlPharmacy, 341

Data enhancement, 3]Data mining, 18, 105-106,207-210,

209-212 (exhibits)Database integration, 3]0-312, 311

(exhibit)Database maintenance, 55-56, 73

address standa.rdization, 6]coding, source/promotional offers,

65-66,65 (exhibit)contact information change, 59-61credit risk/fraud identification, 67de-duping CUstomer files, 57-58, 57

(exhibit)field updating rules, 67-68, 68

(exhibits)householding Customer files, 58-59match coding identification, 62-64, 63

(exhibit)merge/purge processing, 64-65name removal, CUStomer request, 61-62purging records, 59salting files/decoy records, 66~67

scheduling standards, 72storage/security issues, 70-72, 71

(exhibit)summary/aggregate data and, 68-70,

69-70 (exhibits)See also Outsourcing database-

managementDatabase management systems (DBMS),

81,83 (exhibit)Database marketing, 13, 350-351

b-ro-b databases, 364cannibalism and, 164-165consumer databases, 362, 364cost issues in, 16-17database typology, 41ethical issues in, 357~360, 358

(exhibit)global business environment, 17,

351-357,352 (exhibit)industry organizations, 360-362,361

(exhibit)negative perceptions of, 18nor-for-profit databases, 364-365regulation of, 357, 358-359retail competition, 17retailer databases, 365service organization databases,

365-366transaction flow in, 13,14 (exhibit)trends in, 362-366See also Aggregate marketing;

Marketing database developmentDatabase technology, 75-77

computer networks, 77-78, 77(exhibit)

data analysis, 85-86database design, organizational issues,

86-87database development, 91-93database management systems, 81, 83

(exhibit)hardware, 78-80internal needs assessment, 89-91outsourcing, provider selection

process, 88~93software, 80-82, 81-82 (exhibits)structured vs. relational databases,

83-85,84-85 (exhibits)Dell Computer Corporation, 351Demographic data, 32, 46 (exhibit),

47-49,48 (exhibit)cannibalism and, 164-165life-stage market segmentation,

136-138See also Psychographic data; RFM

(recency, &equency, andmonetary) data

Subject Index

391

Direct marketing, 4brand positioning and, 33customer-business dialogue, 6-7profitability tracking in, 10-11sales revenues in, 4-5, 10targeted messaging in, 7See also Database marketing; Internet

database marketingDirect Marketing Association (DMA), 4,

61-62,334e-mail marketing policy, 361-362functions of, 360Privacy Promise, 350, 357, 360, 361

(exhibir)sweepstakes helpline, 359

Distribution structures, 8~9, 11,34Diversification strategy, 31-32Driver's Privacy Protection Act, 45Dry testing, 31

Ecornmerce. See Internet databasemarketing; Internet markers; Onlinecustomers

Electronic Retailing Association (ERA),360-361

E-mail marketing, 321applications in, 321-322formats for, 322-323, 323-324

(exhibits)marketing tests and, 347-348opt-out policy, 361-362See a/so Internet database marketing;

Online customersE'Mail Preference Service (EMPS),

62,362Encryption techniques, 70~71Enterprise Mine); 186, 207, 208~209,

209-211 (exhibirs)Ethical/social concerns, 357~360, 358

(exhibir)Eximious, 203Expected profit calculations, 226

calculation process, 228~230, 228(exhibir)

expected monetary value, 226-227logistic regression model and, 227·228promotion decisions, 230-231

Fair Credit Reporting Act, 45, 67Financial planning, 11Fingerhut Companies, Inc., 337First Union Bank, 78-79Frozen files, 98, 105Fulfillment database, 41, 43

Gains charr. See Response gains chartGeneral Electric, 26

General Foods, 8Geo·demographic data, 47-49, 48

(exhibir),136-:I38Global business environment, 17,

35"1-352,352 (exhibir)competition, evaluation of, 355cultural considerations and, 352-353economic elements, 355human capital issues, 356legal practices and, 353-354media availability, 356-357natural environment and, 356political environment and, 356psychographies and, 354-355sociocultural aspects and, 353technological considerations, 356

Global Online Retailing Report, 313GoldMine, 12, 81, :133

Home Depot, 8Honesty in Sweepstakes Act, 358-359Hot-line lists, 50-51Hypothesis tests for significance, 268

business decision-making and, 279confidence levels/intervals in, 279, 280direction of, 270, 271 (exhibit)error rate, determination of, 269-270hypothesis in, 269p value of, 279-280two sample means, difference between,

270-274two sample proportions, difference

between, 275-278

IBM, 6, 78Info Base, 50Infomercials, 12Information dissemination, 12-] 3

database design and, 86-87retail environment, 13, 15 (exhibit)See also Marketing communicaricns

INSOURCE, 50Integrated Marketing Technology (IMT),

69-70Interactive marketing. See Direct-

marketingInter-Continental Hotels and Resorts, 35 JInternet database marketing, 309-310

m-commerce, 310, 320characteristics of, 314-317commerce, growth in, 312-314cookies and, 320, 330-331customer-initiated databases, .362,364customer service component in,

313-314,315,316database integration in, 310-312, 31 'I

(exhibir)


e-mail marketing, 321-323, 323-324(exhibits)

limitations of, 317-319marketing web sites, assessment of,

334-337,336 (exhibit), 338-340(exhibits)

personalization strategies, 319-321privacy issues in, 320, 354See a/so Online customers

Internet Direct Marketing Bureau(IDMB),361

Internet markets, 7, 8~9banner ads and, 134,318

product information and, 12-13profitability and, 26regulation of, 18security issues, 70-72, 71 (exhibit)

Kellogg, 8KnowledgeSEEKER, 207

Lifetime value (LTV) analysis. See LTVcalculations

List data, 45-47, 46(exhibit), 49-51LivePerson, 314L. L. Bean, 132Logic counter variables, 113-115, 115

(exhibits)factor analysis in, 115longitudinal variables, 1l7~119, 118

(exhibits)ratio variables, 116-117,116-117

(exhibits)Lotus Approach, 80LTV calculations, 11, 69, 239-240

aetuaVaggregare calculations, 243-244CUStomer sample types and, 247-248discount rate/net present value

calculations, 244-247, 245-247(exhibits)

forecasting and, 248methodologies, 240-244profiles, 241-243, 241-243 (exhibits)

Mail Preference Service IMPS), 62Market development strategy, 31Market penetration strategy, 30-31Market planning. See Marketing database

development; Marketing planningprocess

Market segmentation, 5-6, 32, 129-130cannibalism and, 164-165CHAID (chi-squared automated

interaction detection) analysis,148-153,149-152 (exhibits)

cluster analysis, 158-162, 159-162(exhibits)

corporate-level, 134-135, 135 (exhibit)ethical/public policy issues and,

165-166factor analysis, 153-157, 155-157

(exhibit)life-stage segmentation, 136-138objective, definition of, 130-134, 131

(exhibit), 133 (exhibit)product-line specific, 135-136, 135

(exhibit), 137 (exhibit)product proliferation and, 164promotional intensity and, 163-164promotional product offerings and,

134-136results, overgeneralizarion of, 165univariate/cross-tabulation analysis,

138-145,140-143 (exhibits)See also Statistical analysis; Target

marketMarketing communications, 34-35Marketing databases, 2-3, 4customer management, 10, 41development/maintenance of, 5expense/revenue tracking, 10·11marketplace trends and, 5-13sources of, 3-4See a/so Customer data; Database

maintenance; Database marketingMarketing database development, 22-23

computerized databases, 23-24, 24(exhibit)

corporate strategic planning and,26-27,27 (exhibit)

Customer databases, 24-25strategic planning in, 25-26See a/so Customer data; Marketing

planning processMarketing planning process, 27-28, 36,

37 (exhibit)marketing tactics, 33-36market objectives, establishment of,

29-30monitoring/control, 36situational analysis, 28-29strategy development, 30-32, 30

(exhibit)targeting/product positioning, 32-33

Marketing tactics, 33distribution, 34pricing, 35product promotion, 34-35

Marketing test plan/design, 287alternative approaches, 304-308, 306

(exhibits)banner ads, 345-347, 346 (exhibits)bulk mailing, test/control packages

and, 288

Subject Index393

changes, testing of, 289-290 289{exhibit] ,

e-mail campaigns, 347.348full factorial test design, 290 292 293

(exhibit) "outside list test design, 292-294,

293-294 (exhibits)package element interactions, 290-291

291 (exhibit) ,reverse testing, 288-289sample mean and, 295-297sample proportion and, 298-304sample size considerations, 294-304software for; 304test universe, definition of, 291-292

Marketing test results, 251-252analysis software, 285, 285 {exhibit)breakeven response rate level, 282-284confi~~~_~~~ervalcalculations,

gross vs. net responses and, 281hypothesis tests for significance,

268-280impact factors and, 284multiple comparisons, 281-282

Marketivliner; 207, 209, 210, 211-212(exhibit)

Marketplace trends, 5accountability for expenditures, 10-11business functions, integration of, 11customer sophistication, 12-13distribution structure/power; 8-9information dissemination, 12-13lifestyle/demographic trends, 9-10market segmentation, 5-6media fragmentation, 7service/CRM focus, 6-7technological advances, 11·12

Mass advertising, 34-35Match coding, 62-64, 63 (exhibit)Mcommerce, 310, 328Media marketing, 7, 13Mega Telecom, 187-188, 189Microsoft Access, 80, 81 (exhibit)Mobile commerce, 310, 328Molloy, C, 357MotherNature.com,343Multiple regression modeling, 169-170,

181-182binary response data and, 193-194clone/best customer model, 202-204,

205 (exhibit)data mining, tools/software, 207-210,

209-212 (exhibits)data preparation for; 184-187, 185

(exhibit)gains charting and, 194

linear compensatory measurement. 188-189,189 (figure) ,

logistic regression models 192logit (multiple logistic reg~essionJ

models, 199-200,200 (exhibit)rnarkenng objective, definition of,

182-184,183 (exhibit)model-building (MUSCLE) guidelines

213-215 'model formula, 187model interpretation, 187~192,

188-191 (exhibits)multicollinearity and, I92~193,

194-197, 196-198 (exhibits)neural networks, 206outside list modeling options,

201-204, 205 (exhibit)regression diagnostics, 194~199,195

(exhibit)response models, 202, 205 (exhibit)sample-based modeling, 193sample specifications and, 200-20'1,

201 (exhibit)stepwise regression models) 205~206variable significance and, 197-199,

199 (exhibit)See also Expected profit calculations;

Response gains chart

National Change of Address (NCOA)processing, 60

NetCoalition, 362, 363 (exhibit)Net present value (NPV), 244Neural networks, 206New Economy business, 26Nor-for-profit databases, 364-365

Online customers, 327behavior dara, 330-331conversion/retention rates, 337, 341customer-initiated databases, 362, 364data collection and, 327-332internet users vs. online buyers,

332-334marketing tests, 345-348, 346

(exhibits]marketing web sites, assessment of,

334-337,336 (exhibit), 338-340(exhibits]

registration data, 328-330, 329-331(exhibits]

RFM data elements and, 342source data, 332targeting methods and, 341~345

Oracle software, 81Outsourcing database management,

88-89


internal needs assessment, 89-91proposal formulation, 90selection process, 91vendor search, 90-91

Paradox, 12

Personal selling, 4Plan-alyzer, 285, 304Planning. See Marketing database

development; Marketing planningprocess

Point-in-time customer data, 98-100, 100(exhibit)

PosrMasrerDirect.com, 321Predictive Networks, 344-345Pricing, 35Privacy Promise, 350, 357, 360, 361

(exhibit)PRlZM,49Proctor and Gamble, 9, 12, 132Products:

development strategy, 31knowledge of, 12-13positioning of, 32-33proliferation, segmentation and, 164promotion of, 34-35Profitability cracking, 10-11See also Expected profit calculations

Prospecting databases, 24, 41Psychographic data, 32, 46(exhibir)

cannibalism and, 164-165global business environment and,

354-355life-stage market segmentation,

136-138Purchase histories, 13

Regression modeling. See Multipleregression modeling; Sunpie linear-regression modeling

Reporting. See Strategic reporting/analysisResource allocation, 11Response gains chart, 193~194

bootstrapping and, 225-226cumulative gains chart, 222-223, 222

(exhibit)fails sample and, 231gains reconciliation, 231-233, 232

(exhibits)historical gains falloff chart and,

223-225,225 (exhibit)incremental response gains chart,

2'18-222,219-220 (exhibit)See also Expected profit calculations

Retail transactions:customer feedback, 13, 15 (exhibit)database marketing and, 17, 365

market information flow, 13, 15(exhibit)

store-based, 9-10transaction flow in, 13, 14 (exhibit)

Retail transactions, See also Internetdatabase marketing

RFM (recency, frequency, and monetary)data, 43-44, 98, 130

customer file segmentation and,145-148,146-147 (exhibits)

disadvantages of, 147-148not-for-profit market and, 364-365online/offline customer activity and,

342product-line segmentation and,

135-136Ross-Simons, 71 (exhibit), 72

Sampling. See Analysis sampleSAS, 86, 107, 112, 205Segmentation. See Market segmentationService organization databases, 365-366Service/satisfaction focus, 6-7, 35Simple linear regression modeling (SLR),

170-174, 170-173 (exhibits)coefficient of determination and,

174-176,175-176 (exhibits)statistical background of, 176-178,

178 (exhibit)Spam, 318, 322Statistical analysis:

correlation analysis, 120~127, 121(exhibit), 123-126 (exhibits)

database files and, 86modeling and, 49simple linear regression analysis,

176-178,178 (exhibit)See also Analysis samples; Customer

data analysis; Market segmenta-tion; Multiple regressionmodeling

Stepwise regression models, 205-206Store-based retailing, 9-10Strategic planning, 30

diversification, 31-32market development, 31market penetration, 30-31marketing database development,

25-26product development, 31

Strategic reporting/analysis, 235-236impact studies, 248-249key active customer counts, 236-238key list segment counts/statistics,

238-239,239 (exhibit)list vitality customer statistics, 238,

238 (exhibit)

Subject Index

LTV calculation, 239-248promotional intensity; monitoring of,

249-250Structured query language (SQL), 73,

85-86Sweepstakes promotion, 358-359

Target market, 4,7database marketing and, 15-16online customers, 341-345product promotion and, 34-35strategy developmenrJproducr position-

ing, 32-33See also Market segmentation

Telephone Consumer Protection Act, 62

395

Telephone Preference Service (TPS), 62Tiffany & Company, 79Tracking activities:

coding records, 65-66, 65 (exhibit)marketing programs, performance of,

36mass media and, 13profitability,10-11

Transaction data, 53Transaction flow, 13, 14 (exhibit)Wal-Man, 8, 313

Weber-StephensProducts Co" 132

XactMail,321

About the Authors --------

Ronald G. Drozdenko, Ph.D., is Professor and Chair of the MarketingDepartment, Ancell School of Business, Western Connecticut StateUniversity. He is also the founding Director of the Center for BusinessResearch at the Ancell School. He has more than 25 years of teaching expe-nence. The courses he reaches include Strategic Marketing DarabasesInteractive/Direct Marketing Management, Product Management:Marketing Research, and Consumer Behavior. He is collaborating with rheDirect Marketing Educational Foundation to develop a model curriculumfor universities pursuing the areas of interactive or direct marketing.Working with an advisory board of industry experts, he codeveloped theMarketing Database course in the model curriculum. He has codirectedmore than 100 proprietary research projects since 1978 for rhe marketingand research and development departments of several corporations, includ-ing major multinationals. These projects were in the areas of strategic plan-ning, marketing research, product development, direct marketing, and mar-keting database analysis. He also has published several articles and bookchapters. He holds a Ph.D. in Experimental Psychology from the Universityof Missouri, He is a member of the American Marketing Association, theSociery for Consumer Psychology, and the Academy of Marketing Sciences.He is also the coinventor on three U.S. patents.

Perry D. Drake has been involved in the direct marketing industry forapproximately 15 years. He is currendy the Vice President of Dtake Direct,a database marketing consulting firm specializing in response mod~l~ng,customer file segmentation, lifetime value analysis, customer profilll1g,database consulting, and market research. Prior to this, Perry worke~ forapproximately 11 years in a variety of roles at The Reader's Digest

Association.Perry's initial position at The Reader's Digest Association was as a ~tat-

istician in the quantitative analysis department, applying segmentatl?n,response modeling, test design, and multivariate techniques.. Late.r,mov~nginto a product line role as Associate Director of Magazme Clrcu.l~t~onMarketing Perry assumed full strategic responsibility for all acqUisIfIOnefforts including mailings to house and outside lists as welJ as renewal andbilling' efforts. More recendy, Perry assumed responsibility for creaDng

397


a new marketing services division in preparation for the new marketingdatabase, systems, and procedures being implemented at The Reader'sDigest Association. As Director of Marketing Services, Perry was responsi-ble for a staff of over 40 marketing database professionals in support ofmarketing efforts for the entire U.S. business.

In addition ro consulting, Perry has taught at New York University in theDirect Marketing Master's Degree program since Fall, 1998, currentlyteaching "Statistics for Direct Marketers," "Database Modeling" and"Advanced Database Modeling" to future direct marketers. Perry was therecipient of the Center for Direct and Interactive Marketing's "1998-1999Outstanding Master's Faculty Award." This honor was awarded to Perryin recognition of performing at a level above and beyond NYU's standardof quality by providing exceptional academic services to both students andthe program. Perry also lectures on testing and marketing financials forWestern Connecticut State University's Interactive Direct MarketingCertificate Program. Along with Ron, he is collaborating with the DirectMarketing Educational Foundation to develop a model curriculum for uni-versities pursing the area of interactive or direct marketing.

Perry earned a Masters of Science in Applied Statistics from the Universityof Iowa and a Bachelor of Science in Economics from the University ofMissouri. He is a member of the Direct Marketing Association, the DirectMarketing Club of New York, and the American Statistical Association.

445 texbook 0003

Documents

Transcript of 445 texbook 0003