todays ad laws will they survive the digital revolution.pdf

8/2/2019 todays ad laws will they survive the digital revolution.pdf

1/8

Today's Advertising Laws:Wiii They Survive the Digitai Revoiution?

BYRON SHARPEhrenberg-Bass Institute

for Marketing ScienceUniversity of Soutii

AustraliaByron.sharp

marketingscience.infoYORAM (JERRY) WINDTiie Wiiarton SchoolUniversity of [email protected]

Even advertising has scientific laws, empirical patterns that generalize across a widerange of known conditions. These empirical generalizations provide us withbenchmarks, predictions, and valuable insights into how the digital revolution mayaffect advertising. More than ever we need systematic research to understand thegeneralizability of our research findings.

INTRODUCTIONThis special issue of the Journal of AdvertisingResearch {JAR) on empirical laws is based on aninvitation-only conference at held at the WhartonSchool, University of Pennsylvania in December2008. The conference brought together over 100thought leaders from industry and academia topresent and discuss empirical generalizations abouthow advertising and media work. The conference(www.futureofadvertising.wordpress.com) and thisspecial issue of JA R are studies directed by the"Future of Advertising" project of the SEI Center(www.seicenter.wharton.upenn.edu) at the Whar-ton School.

A SPECIAL ISSUEThings were complicated enough for advertisersand advertising researchers when, almost over-night, we found ourselves in the middle of adigital revolution. The fragmentation of tradi-tional media is a great challenge for advertisers.On top of this, we have a range of new media,and, just as importantly, we are seeing radicalnew models of advertisingdifferent ways thatconsumers can interact with advertising, new tar-geting opportunities, delivery of totally differentaudience profiles, different capacities for reach inboth time and space, and so on.

Empowered consumers have far more controlover how they receive their entertainment andnews-and, hence, their advertising exposure. Ourmeasurement systems are struggling to catch up.

while, at the same time, technology is makipossible the collection of metrics that previouwe could only dream of having.

Advertising management today means mumore than commissioning content for commcial media. Advertisers now have to embrace art of influencing (rather than controlling) mesuch as word of mouth, social networks, prodplacement, and a variety of other new advering opportunities.

These changes bring opportunities. But, in ding so, they put incredible pressure on advere r s . It is so much easier to get things wrong nothe risks of what senior managers call "losmoney on advertising" are climbing ... even whadvertising productivity was not high to swith. We can either stumble b lindly into this branew world, or we can take stock of what really know and seek to apply this knowledgethe changes we see around us.

The development of an inventory of substangeneralizab le findings is useful not only from point of view of advancement of science in maring, but also from the point of view of advanmanagerial applications... . (Bass an d W ind, 1

In this special issue, our focus is on fundamtal knowledgeempirical generalizations with l"use-by" dates that hold across a wide rangeknown conditions. Our aim is to collect a listwhat is currently known with some certainty.


2/8

ADVERTISING LAWS AND THE DiGiTAL REVOLUTION

Empirical generalizations then are the building blocksof scientific explanatory theory, and history has taughtus tha t theory not built on empirical laws most o ften

g 1 environment, providing them with

The future is always unc harted, but one

ps between things that have been

If you heat water to 100C, it boils. This

across countries. Indeed this law willtoday in millions of kitchens through-

The predictive value of any law is ex-

the way it is. Empirical generalizationsthen are the building blocks of scientificexplanatory theory (Bass, 1993), and his-tory has taught us that theory no t built onempirical laws most often turns out to bewrong. The world is astonishingly com-plicated and counter intuitive, so arm-chair theorizing regularly lets us down.

Without the bedrock of scientific laws,deep thinking by the best and m ost learnedminds h a s an unproductive tendency t o p r o -duce elaborate theories that a r e plain wrong.Western medicine for nearly 2000 years builta complex edifice of theory based on theidea of humorial imbalance, intermingledwith astrology (another theoretical endeavorthat ignored the need for empirical laws).The results for practice were terrible, a s any-one unfortunate enough to be attended toby a medieval doctor can attest. Doctors,bleeding patients to balance their humo rs,caused countless deaths.EVEN ADVERTISING HAS LAWS

. Contrary to popular belief, the socialsciencesincluding marketinglendthemselves to empirical laws (Ehrenberg,1993). For example, the reach of an adver-tising campaign increases with spend, b utnot monotonicallyso the reach returnon spend diminishes, quite rapidly after apoint. This is true for a wide range ofconditions, including different m edia. Thispattern is fundamental practical knowl-edge for anyone planning an advertisingcampaign (yet such core knowledg e is notalways taught in marketing courses).

Underlying this empirical generalization is another about viewing rates having highly skewed distributionsa fewheavy viewers and a "long tail" of lighterviewers. New media appear to show thesame predictable distributions (Park andFader, 2004). Laws like this shine a lighto guide us through the digital advertising revolution.

It is a common misconception that scientific laws in the physical sciences aredifferent, that physical laws are somehow absolute and inviolate. Nobel laureate Herbert Simon (1968) eloquently pointsout that any chemistry student can easilyshow violations to Boyle's law, and thaanyone with a rock and a feather can"disprove" Galileo's law of falling bod-i e s , showing that physical science lawsare also approximate. As empirical ac-counts of deviations grow, we refine andlimit the conditions under which a lawapplies. Einstein's theory of relativity didnot mean that New ton's laws were wrongthey just do not apply at velocities ap-proaching the speed of light. It is naturalthat we gradually refine a law, or replaceit with one with similar accuracy, buteven greater generalizability.

Scientific laws, even in the physical sci-ences, are approximate. One of the keypurposes of science is to simplify and,therefore, understand our complex world.So scientists are more interested in a sim-ple law/model that is known to hold (al-beit approximately) over a wide range ofconditions than a complex model that fitsvery well, but to a very limited range ofconditions.

Advertising laws are not different, theyare just fewer and less developed than w emight like. One has to look to find laws,and look further to refine them. And, inthis regard, marketing researchers have apoor track record. We waste a lot of re-search by treating each study as "standalone" (Ehrenberg, 1966). Even worse.


3/8

ADVERTISING LAWS AND THE DIGITAL REVOLUTION

when laws exist we tend to overlook them,consistently trying to formulate newmodelseach one in ignorance of the es-tablished empirically grounded knowl-edge. Funds for marketing R&D arelimited; we cannot afford to keep makingthese mistakes (see Uncles and Wright,2004, for suggestions how to avoid thesemistakes).

A solution is to take a multiple-sets-of-data approach. And even if we have asingle set of data, we can slice it up basedon known differences (e.g., men com-pared with women, print compared withTV advertising exposure, early-eveningviewing compared with late-evening view-i n g ) . We then can start looking for pat-terns that generalize across the conditions.If we fit a single model to the one dataset, we not only lose this insight, but wealso lose the opportunity to discover amodel tha t generalizes across many or allof the conditions we have studied.

Traditional statistical training in busi-ness schools almost always teach a "best-fit" modeling approach rather than seekingto uncover a model that fits well overdifferent conditions. This needs to changeif we are to uncover more m arketing laws.WHAT ARE EMPIRICALGENERALIZATIONS?"Empirical genera lizations" is a fancy termfor what also are known as natural lawsor scientific laws, events that occur inrepeating patterns.An example: Many thousands of yearsa g o , our ancestors noted that the pin-pricks of light in the sky at night movedthroughout the evening. And they no-ticed that this happened every evening.Further study showed that they all movedtogether, except for a few that did not.(Ancient Greeks called them "steres plan-etai" or "wandering stars." We now callthem them "planets.") Yet more study re-vealed an an nual pattern. Over time, these

empirical patterns were well documented,and it became possible to predict where astar would be at any time of the year.These incredibly practical empirical gen-eralizations became the basis for naviga-tion for many years. Eventually thepatterns were distilled into a heliocentricmodel that used the earth's spinning toexplain the stars (and our sun's)"movement"a very counterintuitive ideaat the time, but one that fits all the em-pirical patterns that had been collected.And, of course, without these laws, wenever would have arrived at this radicalnew world view.

It seems trivial to point out that ourancestors also noted that the predictablemovement of the stars happened in goodand bad weather, during peace and wartime, and so on. But this is very impor-tant because it reveals what the patternsdepend on and, thereby, gives us insightinto what causes what, what is important,and what is irrelevant. Kings or priestscould pass edicts, but they could not stopthe stars following their law-like path.

Of course, as knowledge accumulated,people began to refine their findings: Therising and setting of stars vary de pendingon the time of year and where you arestandingknowledge that was useful foragriculture and for navigation. And a vi-tal part of any empirical generalizationremains documenting where a patternholds and where it does not; wha t circum-stances matter and which do not.LAWS AND THEORYHistorically, the terms "theo ry" and "law"are used interchangeably. Whether a set oflaws given one title or the other app ears tobe simply due to historical accident, hencewe refer to "Newton's laws" and "Ein-stein's theory." The marketing philoso-pher of science Shelby Hunt, however,defines a (well-developed scientific) theoryas a set of systematically integrated laws

(Hunt, 1991). This is an important poiseveral laws can "fit tog ether" to give a nworld view.

Darwin's theory of natural selectifor example, essentially is a story told three laws that h e scribbled into his nobook in 1838 (20 years before the publition of his Origin of the Species):

Three principles will account for all:

1 . grandchildren, like, grandfathers2 . tendency to small change ... especia

with physical change3 . great fertility in proportion to supportparents.

Darwin realized that three law-like pterns (inheritance of traits, with slight rdom mutatio ns, and the fact that all specproduced far more offspring than wousurvive) together told a story of comption for survival, natural selection, athe evolution of species (see Darwin aQuammen, 2008).

So empirical generalizations rapidly come deeply theoretical. But theory tis not built on laws seldom lasts the of time.

When a law-like pattern is discoveredinstantly begs an explanation why the geralization happens to fit the facts. Texplanationthis proposed mechanismthen is tested by the search for particuboundary conditions. T h e explanation neto hold up as conditions are shown noaffect the law and boun dary conditionsdiscovered. For example, an explanationmarketing's famous Double Jeopardy cannot be particular to a prod uct categfor it has been shown to hold for hu ndrof categorie s. So an exp lanatio n of this has to apply to brand choice, attitudes, TV viewing, to name just a few of the geralized conditions.

If the empirical generalization is "rent users of a brand are twice as likel


4/8


5/8


2I1 0 0 %

Penetration of the MarketFigure 1 GraphicalRepresentat ion ofMarketing's DoubleJeopardy Law.theories, folklore, anecdotes, or even gos-sip can send us off looking for a pattern.Many of advertising's theories, checklists,and rules of thumb have this sort of basis.This is fine so long as these principles arethen empirically tested and eventuallyturned into empirical laws, otherwise mar-keting practice risks being like medievalmedicine.

WHAT EMPIRICAL LAWS LOOK LIKEEmpirical laws can often be expressed as"if/then" type statements. For example,marketing's Double Jeopardy law (Ehren-berg, Goodhardt, and Barwise, 1990) saysthat larger brands have more customersand those customers are slightly moreloyaL This can be expressed as, "//a brandhas higher market share than a rival, thenit will have a larger customer base that isalso slightly more loyal" or ";/ two rivalbrands have similar market share, thenthey will also have similar sized customerbases." A statement like either of theseeasily can be turned into a normative "dothis" statement: "If you want to substan-tially grow your market share, then youneed to get more customers."

Empirical laws can also be expressedalgebraically or shown graphically, e.g..Double Jeopardy as w = wo/{\ - V ) or asshown in Figure 1.

THE NEED FOR MULTIPLE SETS OF DATAA scientific law: must be about the real world (in this

case, advertising), not just a logical state-ment (i.e., "sales revenue rises if youmake more sales") (Hunt, 1991);

must have been observed a number oftimes;

should have been observed across awide variety of conditions.

We must be able to say something aboutthe conditions under which it generalizes,and the conditions under which it doesnot.

We can discover interesting things witha single set of data, but, by itself, a singledata set cannot really tell us anyth ing abo utthe generalizability of our finding. W e don'teven know if we'll get the same result ifwe immediately do the study again.

It is a common mistake to think that sta-tistically significant results are generaliz-able, when actually all these significancetests tell us is that our discovery probablyis not merely an artifact of random sam-pling variation. A significance test doe s nottell us the conditions unde r which the lawdoes or does not hold; we are still just guess-ing about which population/circum stancesour finding generalizes. Hall and Stamp(2004) made this mistake in their other-wise worthy attempt to catalog empiricalgeneralizations . Meaningful Marketing: 100Data-Proven Truths.

W e always learn something from repeat-ing a study on another set of data. A per-fect replication must, by definition, give usthe same finding and so tells us nothing. Aperfect replication is practically impossi-b l e , however, so a very close replicationgives us, at the very least, a bit more con-fidence that the original result was not dueto some undocum ented error orflukeevent.If the finding holds firm across deliber-ately manipulated conditions (e.g., differ-

ent researchers, different product categodifferent country, different commercials, ferent media, different segments), then not only gain confidence in our finding, we also learn about its generalizabilityand this is the key to prediction and deexplanation.THESE ARE NOT EMPIRICAL LAWSIt has been observed that many of supposed generalizations in marketpractice (Cierpicki, Wright, and Sha2000) and literature are "either tautogies, truisms, or so overly general tthey are of very limited use in develing marketing science" (Halbert, 19p . 66). It is, therefore, worth pointing the sort of statements that appear lalike, but fa on essential criteria (see H1 9 9 1 , p. 105 for further examples adiscussion): "If God is not happy with the adver

ing then it will not work" is not a valaw. Because God is not empiricalwe cannot check if He really wa s hapor not.

Tautological statements (e.g., "the vertising will lift sales or not") are empirical generalizations because tare true simply by logic and so do really say anything about the real wo

The law-like statement "trial pchase occurs only if the consumer psesses a sufficiently favorable attitu(Rossiter, 1987) is rendered tautolcal and unfalsifiable by the w"sufficiently"because whether unobservable attitude is judged sucient depends on whether the csumer made the trial purchase. So statement becomes meaningless. "If don't allocate adequate media wethen your advertising will fail" is silar, tmless "adequate" can be spfied independently of campaign succor failure.


6/8


THREE COMMON MISCONCEPTIONS ABOUTSCIENTIFIC LAWSEmpirical laws don't tell a manager what to doMarketing is a professional practice and managers have goals they want to achieve.Thus there is added value in taking a descriptive law and expressing it as a normativelaw. For example, if the empirical generalization is that advertising in multiple mediagerierates higher sales than the same dollars spent on a single media, then we caneasily craft a normative statement "to generate higher sales spread your advertisingdollars over multiple media." All normative laws contain an assumption about a goal(wfiether or not it is explicitly stated).Empirical laws are of no value unless they are turned into prescriptive statementsNormative statements are useful, and so are descriptive statements because theyprovide benchmarks and norms. Newton's laws are purely descriptive, but of greatpractical value to any engineer, pilot, or rocket s cientist. Anyone familiar with marketing'sDouble Jeopardy law (Ehrenberg, Goodhardt, and Barwise, 19 90) can p redict the level ofrepeat purchasing a brand will obtain if it reaches a particular level of market share. Atfacje value this might seem disappo inting because the law does not say how to get tha tlevel of market share. But this is unreasonable: Newton's laws don't say how to build arocket either. There is plenty of room of creativity and engineeringfor both rocketscientists and marketers.Managers should try to beat the lawEmpirical laws describe the way the world is. They help us explain and p redict. They areto be used, not beaten. In physics, escape velocity is a consequence of the law ofconservation of energy.^ Rocket scientists don't seek to beat this law; they use it todesign rockets that will actually work.'Perpetual motion m achines try to beat this law. Patent offices routinely receive patent requests for such devices thoughnone have been built. W hile no scientific law is absolute and a bounda ry condition ma y be discovered one day, readerswould be prudent not to invest in business ventures based on building a perpetual motion machine.

become arrogant" is similar. The prob-lem with this (the 19th "ImmutableLaw," according to Ries and Trout, 1994)is that the conditions are not specified;of course you could become arrogant,or not, whether you are successful ornot. So this is not an empirical lawbecause it really says nothing.

I

The key point is that every empirical

potential for some

wrong and needs to be amended. Theabove examples fail because it is impos-sible to ever prove them wrong with em-pirical data.WILL OUR ADVERTISiNG LAWSSURViVE?Empirical generalizations are based on his-torical data. Given the dramatic changesin the business environment, the questionarises regarding the value of our currentempirical laws and their applicability andrelevance to the changed conditions oftoday and the uncertain conditions oftomorrow.

The more we know about the general-izability of an empirical law, the betterequipped we are to answer such ques-tions. This points up the value of replica-tion and extension studies, deliberateresearch attempts to see how far lawsgeneralizea call that has been mademany times previously (Hubbard andArmstrong, 1994).

There are some empirical generaliza-tions that have not only stood the test oftime, but they have been examined by dif-ferent researchers, using different meth-o d s . They are known to hold for differentmedia, in different countries, and so on.Such robust laws are more likely to keepon working. (This reasoning might be in-ductive, but it needn't necessarily be na-i v e . If we know what factors the law isimmune to, and we know what factorsare changing in the world, then we canmake intelligent evaluations of whethe r ornot the law will continue to apply.) For ex-ample, in this Special Issue, we have lawsconcerning TV viewing that have held fordecades across a wide variety of condi-tions. Accordingly, Sharp, Beal, and Col-lins (p. 211) argue, in spite of the manychanges wrought by the digital revolutionT V , that advertising will continue to retainmuch of its power. Similarly, Rubinson( p . 220) pulls together a diverse range ofstudies examining the effectiveness of TVadvertising and finds there is no evidencethat TV advertising is losing effectiveness.Jamho uri and Winiartz ( p . 2 2 7 ) , across manycountries, examine consumer perceptionsof the impact of advertising on them andreport a related conclusion. Binet and Field( p . 130) find that cam paigns that include dTV in their mix outperformed those thatdid not.

WHAT NEXT?While the array of laws presented inthis Special Issue is exciting and of im-mense practical value, it is immediately

J u n e 2 0 0 9 J D U R O fl L O F H D U E R T I S II 1 G R E S E H R C H 1 2 5


7/8


apparent that not enough is known abouteach law's generalizabilitywhere it doesand do es not hold. We need m ore system-atic research across a wide range of con-ditions, at very least different media. Weneed R&D to examine which laws aresurviving the digital revolution.

And we need more laws.Not all our scientific laws of advertis-

ing may survive the digital revolution,but they may help us to survive it!BYRON SHARP is a professor of marketing scienceand director of the Ehrenberg-Bass institute atthe University of South Australia. Hisresearch isfunded bycorporations around theworid inciudingCoca-Cola, iVIars, P&G, Kraft, Turner B roadcasting,CBS, and theAustraiian Research Council. He iscurrentiy writing a book. Laws of Growth (seewww.MarketingLawsofGrowth.com).

YoRAM ( " JERRY") WIND isThe Lauder Professor, aprofessor of marketing, the director of the SEi Centerfor Advanced Studies inManagement, and academicdirector ofThe Wharton Fellows Progrann at the Uni-versity of Pennsyivania. He currentiy is leading the"Future ofAdvertising" project that has overseen thisSpecial Issue (http://seicenter.wharton.upenn .edu/).

REFERENCES

B A S S , FRANK M. "The Future of Research inMarket ing; Ma rket ing Sc i ence ." Journal of Mar-keting Research 30 , 1 (1993): 1-6.

, and JERRY W IND. "Introduction to theSpecial Issue: Empirical General izat ions in Mar-ket ing ." Marketing Science 14, 3, P a r t 2 (1995):G l - 1 9 .

C I E R P I C K I , S T E V E N , M A L C O L M W R I G H T , and B Y -RON SHARP. " M a n a g e r s ' K n o w l e d g e of Market -ing Pr inc ip l es : The C a s e of New P r o d u c tD e v e l o p m e n t . " journal of Empirical Generalisa-tions inMarketing Science 5 (2000): 771-90.

D A R W I N , C H A R L E S , and D A V I D Q U A M M E N . Onthe Origin of Species: The Illustrated Edition. N e wYork: Sterl ing Publ ishing Co. Inc. , 2008.

EHRENBERG, ANDREW S. C. "Laws in Market-ing: A Tail-Piece." Applied Statistics London 15,November (1966): 257-67.

. "Even the Socia l Sc i ences Have Laws."Nature 365, 30 (1993): 385.

, G E R A L D G O O D H A R D T , and T. PATRICKBARWISE. "Double Jeopardy Revi s i t ed ." Journalof Marketing 54 , 3 (1990): 82-91.

HALBERT, M . The Meaning and Sources of Mar-keting Theory, T h e o ry ed. New York: McGraw-Hill, 1965.

H A L L , D O U G L A S , and JEFFREY STAMP. Meaning-ful Marketing: WO Data-Proven Truths and 402Practical Ideas for Selling More with Less Effort.Cincinnat i , OH: Emmis Books , 2004.

HuBBARD, R A Y M O N D , and J. S C O T T A R M S T R"Repl i ca t ions and Extens ions in M a rk eRare ly Publ i shed but Q u i t e C o n t r a r y " Inational Journal of Research in Marketing 11233-48.

HUNT, SHELBY. Modern Marketing Theory: Ccal Issues in the Philosophy ofMarketing SciCincinnat i , OH: South-Western Publ i sh ing1991.

P A R K , Y O U N G - H O O N , and P E T E R S . FADER. "Me l ing Browsing Behavior at M u l t i p l e Wsi tes." Marketing Science 2 3, 3 (2004): 280-3

RiES, AL, and JACK TROUT. The 22 ImmuLaws of Marketing: Violate Them at YourRiskl New York: HarperBusiness , 1994.

ROSSITER, JOHN R. "Comments on 'ConsuBeliefs and B ra n d U s a g e ' and on E h re n bATR Model . " Journal of the Market Researcciety 29 , 1 (1987): 83-93 .

S I M O N , H E R B E R T A. "On Judging the Plauit y of Theor i es . " In Logic, Methodology andlosophy of Sciences HI. Amsterdam: New HoPubl ishing, 1968.

U N C L E S , M A R K D., and M A L C O L M W R I"Edi torial : Empirical General isat ion in Maing." Australasian M arketing Journal 12 , 3 (25-12.

1 2 6 J O U R n H L o r f l D U E R T I S l O G R E S E H R C H J u n e 2 o o g


8/8

todays ad laws will they survive the digital revolution.pdf

Documents

Transcript of todays ad laws will they survive the digital revolution.pdf