Part F Innovation and New Products - Amazon...

2015 AMA Winter Educators’ Proceedings F-1

Part FInnovation and New Products

Track ChairsKersi Antia, Western University

Deepa Chandrasekaran, University of Texas at Austin

Innovation Processes and OutcomesAn Empirical Investigation of Composite Product Choice F-2

Kalpesh Desai, Dinesh Gauri, Yu MaThe Effect of Superstar Software in the Video Game Industry: The Moderating Role of Product Generation

Lifecycles F-12Richard Gretz, Suman Basuroy

The Mitigating Effect of Personal Agency on Escalation of Commitment F-13Sunil Contractor

An Exploratory Study of Antecedents and Consequences of Radical Product Innovation Capability F-15Sanjit Sengupta, Stanley F. Slater, Jakki J. Mohr

Open, Low-Cost InnovationAvoiding a Babylonian Confusion: A Systematic Review on Low-Cost Innovation F-17

Ronny ReinhardtDevelopment of Successful Really New Products: The “Over-Collaboration” Effect at Different Stages of the

New Product Development Process F-18Johannes S. Deker, Monika C. Schuhmacher, Sabine Kuester

Sustainability, Open Innovation, and New Product Program Success F-20Shuili Du, Ludwig Bstieler, Goksel Yalcinkaya

It’s All Your Fault! Attributing Blame for Co-Created New Product Failures in B2B Relationships F-22B.J. Allen

IntroductionA currently popular marketing strategy for increasingbrands’ attractiveness to consumers is the introduction ofcomposite products comprising complementary brandedcomponents. Examples include Downy fabric softener inTide laundry detergent, Cascade automatic dishwashingdetergent with Dawn dishwashing liquid, Andes candy inPerry’s ice cream, and Lays potato chips with KC Master-piece vinegar. This composite product strategy uses onebrand from each of the two complementary product cate-gories as a component in the composite product. As anexample to clarify our terminology, the product incorporat-ing Downy fabric softener in Tide laundry detergent isreferred to as the composite product, Tide is the primarybrand, laundry detergent is the primary category, Downy isthe secondary brand,1 and fabric softener is the secondary

category. Tide and Downy are jointly referred to as partnerbrands, and reciprocal effects refer to the influence of thechoice of composite product on the boost in attitudes towardprimary and secondary brands. The strategies of ingredientbranding and composite products differ slightly. In ingredi-ent branding, one brand is a component or part of the finalproduct that is sold under a different brand name (e.g., Intelin an HP computer, Nutrasweet in Diet Coke). In the case ofcomposite products, two brands (primary and secondary)from complementary categories act as components in thecomposite product. However, we rely on the ingredientbranding literature to guide our theorizing, because the lit-erature regarding ingredient branding is well established andthe difference between these two strategies is minor.

Prior research in ingredient branding has investigated thedistinct roles that a secondary brand plays in the primarycategory. For example, experimental data showed how aningredient co-brand (Godiva) can help a primary brand(Slim-Fast) overcome a limitation (poor taste) and improve

An Empirical Investigation of CompositeProduct ChoiceKalpesh Desai, University of MissouriDinesh Gauri, Syracuse University Yu Ma, University of Alberta

ABSTRACTPrior ingredient branding research has examined the influence of “stated” factors such as fit between partner brands on com-posite product (e.g., Tide with Downy fabric softener) attitudes. This research focuses on choice of composite products, andaddresses three managerially relevant questions: Which consumer segments are more likely to adopt the composite product?Will the choice of the composite product have positive or negative reciprocal effects on partner brands? Will the introductionof the composite product benefit the primary or the secondary brand more? The authors use a brand choice model to investigatethe “revealed” choice of complements-based composite products. Study results indicate that (i) despite high fit between thecomposite product and the primary brand, consumer segments have different choice likelihoods for these products, whereasprior research suggests equal likelihood; (ii) the choice of a composite product may not provide a positive reciprocal effect tothe secondary brand; and (iii) the introduction of a composite product may benefit the primary brand more than the secondarybrand, whereas prior research suggests a symmetrical benefit for the partner brands. Finally, the finding that introducing a com-posite product may not cannibalize the sale of the primary brand extends the ingredient branding literature, which has beensilent on this issue.

Keywords: composite product, ingredient branding, choice models, scanner panel data

For further information contact: Dinesh Gauri, Syracuse University (e-mail: [email protected]).

F-2 2015 AMA Winter Educators’ Proceedings

1An alternate terminology could be host brand for the primary brandand ingredient brand for the secondary brand, but for clarity, we useprimary and secondary brand. We thank an anonymous reviewer forthis suggestion.


its evaluation when extending into a new category (Slim-Fast cake mix) (Park, Jun, and Shocker 1996). Similarly, therole of the secondary brand may be to lend a specific char-acteristic to the primary brand (e.g., Irish Spring bath soaplending its scent to Tide laundry detergent) or to provide theprimary brand with a new attribute (e.g., Dayquil addingcough relief to Life Savers candy) (Desai and Keller 2002).

While prior research has enhanced understanding of theingredient branding, these earlier studies focused primarilyon stated factors (e.g., reported brand attitude and statedbrand familiarity) that influence attitudes toward the ingre-dient product and partner brands. Controlled lab experimentsmay not necessarily reflect actual consumer behavior andhence may limit the external validity of the findings (Gey-lani, Hofstede, and Inman 2008). Although many new prod-ucts combining complementary offerings have appeared onthe market, the market structure literature contains littleresearch on these products (Elrod et al. 2002). Thisexploratory study investigates consumers’ revealed choiceof these products by employing scanner panel data to exam-ine important questions of interest to retailers and brandmanagers. The results not only fill key gaps but also answerrecent calls for more research on ingredient branding (Kellerand Lehmann 2006) and multiple-category decision making(Russell et al. 1999).

Our research addresses three interrelated managerially rele-vant questions. First, which segments of consumers are morelikely to choose composite products (that combine primaryand secondary brands) and thus be targeted by marketers?Theoretically, might any factors hold back a consumer fromchoosing the composite product despite currently using (andliking) either or both partner brands? Second, will choosingthis composite product positively or adversely change theadopter’s evaluation of either or both partner brands? Theo-retically, will the consumer’s evaluation of the secondarybrand, post-choice of the composite product, go up becauseit is the only brand in the secondary category that partnerswith the high-equity primary brand, or will it go downbecause the performance of the secondary brand is now lessvisible as part of the performance of the composite product?Third, will the introduction of the composite product benefitthe primary brand more than the secondary brand? Theoreti-cally, in addition to changes in market share because of thecomposite product’s likely cannibalization of the choice ofindividual partner brands, will the introduction of the com-posite product increase the competitive clout or vulnerabilityof partner brands?

Data DescriptionThe data for our study come from a suburban grocery marketof a mid-sized city in the northeastern U.S. For this study,

we obtained access to a dominant local retail chain’s dailytransactions data, which provide information about the dateof the transactions and bonus card holder information(whether a shopper card was used), as well as dollar volume,unit price, quantity, category, and coupon use for every UPCsold. For product stimuli, we selected two composite prod-ucts according to the following considerations. First, thecomposite product should have been introduced somewherein the middle of the three years for which we have the datato facilitate the comparison of pre- versus post-introductionperiods. Second, the composite product should have a rea-sonable level of penetration (we selected only compositeproducts that were bought by at least 1,000 households afterthe introduction). Third, since unbiased identification of theeffects on the primary and secondary brands is important,the composite product should be a combination of only oneprimary brand and one secondary brand. Fourth, the primarybrand should have a prior history of marketing without thesecondary brand to facilitate the comparison of perform-ances of the primary brand in pre- versus post-compositeproduct introduction scenarios. Similar criteria applied to thesecondary brand. Fifth, the secondary brand should be aningredient in only one product, which should facilitate a“cleaner” attribution of changes in the choice of the second-ary brand post-introduction of the composite product. Simi-larly, the primary brand should have only one secondarybrand and should not feature distinct ingredients in differentline extensions of the primary brand.

Two composite products, Tide laundry detergent with Downyfabric softener and Cascade automatic dishwasher soap withDawn dishwashing liquid, satisfied our criteria. Tide (pri-mary brand) belongs to the laundry detergent category, whileDowny (secondary brand) belongs to the fabric softener cate-gory. Tide with Downy represents a complementary compos-ite product because consumers use these products together towash and dry clothes. Similarly, Cascade (primary brand)belongs to the automatic dishwashing detergent categorywhile Dawn (secondary brand) belongs to the manual dish-washing liquid category, and both could be used together forthe same occasion (pre-wash dishes manually with Dawnbefore putting into the dishwasher and cleaning with Cas-cade). We also empirically verify the functional relationshipbetween these two sets of product categories.

Since consumers buy composite products less frequently,our analyses included purchase data for 45 weeks beforeintroduction and 45 weeks after introduction of the compos-ite product (2003–2005).2 We selected only consumers who2We have 45 weeks of post-introduction data for Tide with Downy,which is the shortest time frame among all pre- and post- data peri-ods for both pairs. To run consistent analyses over similar time peri-ods, we used 45 weeks pre- and post-introduction data for both cases.


had made at least one purchase in the primary and secondarycategories before the introduction of the composite product(Swaminathan, Fox, and Reddy 2001). Using this criterion,we were able to have 5,215 and 2,604 households in the Tideand Cascade composite product samples, respectively.

We considered all the brands in the corresponding cate-gories. Tables 1 provide some descriptive statistics about thedata in the period before the introduction of the compositeproduct. It shows that Tide and Downy are market leaders intheir respective categories and that each brand has the high-est price (in cents per ounce) in its category. Arm & Hammerreceives the maximum feature and display support in laun-dry detergent category. Tide and Downy have the largestnumber of SKUs and receive the largest advertising spend-ing3 in their respective categories. Availability of manufac-turer coupons4 is also the highest for both Tide and Downy.

Classification of Households Based on Experiencewith the Primary and Secondary Brands

In the conceptual section, we discussed the classification ofcustomers into three segments based on experience with theprimary and secondary brands. Swaminathan, Fox, and Reddy(2001) defined user experience as the relative purchase fre-quency. We define user experience similarly, except that weuse relative purchase volume instead of relative purchase fre-quency. As our data show a fairly large variation in purchasequantities within the same household over time, purchase vol-ume is a better indicator of consumption experience.

Let PBE and SBE represent the cumulative experience orfamiliarity with the primary and secondary brands prior tothe introduction of the composite products, respectively. Wecompute PBE/SBE as the purchase share of the primary/secondary brand in the first half of the data’s time period,before the introduction of the composite product. For exam-ple, the PBE in laundry detergent is defined as the purchasevolume of Tide/total volume of detergent purchases, prior tothe introduction of the composite product. It follows that the

Table 1. Descriptive Statistics of Brands in Detergents and Fabric Softener Categories Before the Introduction ofComposite Brand

Log Avg. # (weekly Price Coupon of UPCs advt. (cents/oz) Feature Display Availability in a week spending) MarketBrand Share Mean SD Mean SD Mean SD Mean SD Mean SD Mean SDLaundry detergents

Tide 30.14 6.81 .70 .31 .30 .12 .26 .19 .07 24.65 2.54 12.98 2.39Arm & Hammer 18.90 4.12 .84 .81 .30 .75 .36 .05 .08 9.10 1.85 8.94 5.57Era 15.08 4.63 .93 .35 .44 .16 .35 .01 .02 2.94 .31 .00 .00All 10.50 4.62 .34 .49 .43 .11 .29 .06 .05 15.83 1.32 8.91 5.67Purex 7.51 4.05 .70 .40 .42 .32 .42 .15 .14 7.2 .66 1.94 4.07Wisk 6.57 5.38 .58 .46 .40 .21 .34 .12 .09 7.02 .84 1.19 3.21Xtra 5.88 2.10 .22 .38 .44 .24 .40 .00 .00 5.31 .83 .00 .00Cheer 3.47 6.58 .46 .15 .33 .13 .31 .06 .06 5.58 .85 8.31 5.19Private Label 3.26 2.31 .52 .37 .43 .02 .13 .00 .00 4.54 .56 .00 .00

Fabric softenersDowny 36.64 8.95 1.11 .29 .35 .13 .29 .28 .10 23.12 1.45 13.13 .91Private Label 19.81 3.69 .68 .46 .33 .10 .24 .00 .00 13.07 .40 .00 .00Snuggle 13.99 6.62 .46 .48 .37 .14 .28 .16 .08 12.53 .79 8.46 5.90Bounce 14.25 4.80 .17 .24 .32 .00 .05 .23 .13 7.85 .43 7.82 5.49Arm & Hammer 6.73 4.59 .37 .34 .43 .20 .39 .16 .17 3.97 .16 8.97 5.57All 3.41 4.19 .12 .43 .40 .11 .27 .09 .11 4.28 .66 4.34 5.78Cling Free 5.17 3.05 .23 .38 .42 .16 .33 .21 .20 2.00 .00 .00 .00

3The estimated weekly national advertising expenditures of thebrands were obtained from TNS Media Intelligence for the corre-sponding period of the data. 4Manufacturer coupons are coded as 1 if available for selected mar-ket at a given week and 0 otherwise.


brand experience is customer-specific. PBE (SBE) of 1 sug-gests a loyal user of the primary (secondary) brand, whereasPBE (SBE) of 0 suggests the household is a non-user of theprimary (secondary) brand. A value of PBE (SBE) between0 and 1 implies that the household is a familiar user whoswitches between the primary (secondary) brand and otherbrands in the primary (secondary) category.

Tables 2a and 2b show the percentage of users in varioussegments, classified on the basis of their experience with theprimary and secondary brands.5 Of all the households in ourdetergent composite product sample, 12.04% and 11.85%constitute loyal users of the primary brand (Tide) and thesecondary brand (Downy), respectively, while the percent-

age of users that are loyal to both Tide and Downy is rela-tively small (2.42%). Of the households in the dishwashingcomposite product sample, 23.85 % and 24.58% are loyalusers of the primary brand (Cascade) and the secondarybrand (Dawn), respectively. The percentage of users that areloyal to both Cascade and Dawn is small here also (7.07%),although a bit larger than in the other case.

Model SpecificationOur goal is to estimate a shift in consumers’ brand prefer-ence after introduction of the composite product. To isolatethe change in brand preference owing to changes in market-ing activities, we employ brand choice models (Guadagniand Little 1983; Krishnamurthi and Raj 1988; Lattin andBucklin 1989) in the primary and secondary categories, esti-mated simultaneously and incorporating household-levelheterogeneity.

We illustrate the model specification using the example of adetergent and softener composite product. For both cate-gories, we assume that households derive utility from thepurchase and consumption of a brand. Computation feasibil-ity leads to analysis at the brand level (e.g., Erdem 1998;Mehta 2007) instead of the SKU level (Fader and Hardie

Table 2a. Segmentation of Users Based on Prior Primary and Secondary Brand Experience for Laundry Detergent and Fabric Softener Categories

Percentage of Customers Secondary Brand (Downy) Experience Non-users Familiar Users Loyal Users TotalPrimary brand (Tide) experience

Non-users 47.02 7.88 6.12 61.02Familiar users 18.43 5.20 3.32 26.94Loyal users 7.56 2.07 2.42 12.04Total 73.00 15.15 11.85 100.00

Table 2b. Segmentation of Users Based on Prior Primary and Secondary Brand Experience for Automatic and Hand Dish washing Detergent Categories

Percentage of Customers Secondary Brand (Dawn) Experience Non-users Familiar Users Loyal Users TotalPrimary brand (Cascade) experience

Non-users 20.85 16.82 10.02 47.70Familiar users 13.06 7.91 7.49 28.46Loyal users 10.91 5.88 7.07 23.85Total 44.82 30.61 24.58 100.00

5Note that our segmentation of consumers is based on their initialpurchases. This way of segmenting, while facilitating managerialinterpretations, may introduce two potential biases owing to the“regression to the mean” (RTM) phenomenon and ceiling effect. Forexample, some consumers may be classified into the loyalty segmentby pure chance, even if their choice probability of the focal brand isless than 100%. As a result, the introduction of the composite brandmay seem to reduce these customers’ choice probability of the focalbrand owing to RTM. The ceiling effect might also occur—the con-sumers with the highest probability of buying the focal brand wouldhave less room to improve.


1996). Although examining SKU-level choice would revealadditional substitution patterns among SKUs on attributessuch as size and scent, this approach would not enhance ourresearch focus on brand alliance. In addition, we isolate thecomposite product from the primary brand and treat it as aseparate alternative in the primary category so that we cancalculate potential cannibalization within the primary cate-gory without losing tractability.

The utility is composed of a deterministic part and a randomerror with certain distribution assumptions. We assume thatthe deterministic part can be written as a linear function ofthe brand preference, price, and other marketing mixvariables. Households choose the brand that offers the mostutility within the choice set. Formally, let h denote a house-hold, t denote time, X denote a vector of marketing mixvariables (including price, feature, display, number of UPCs,coupon availability, state dependence,6 and advertisingspending), and t0 denote the time when the composite prod-uct was launched. Let P denote the primary category, Sdenote the secondary category. Let j denote a brand in theprimary category, where j=1,…J. Let k denote a brand in thesecondary category, where k=1,…,K. Furthermore, let cpand pb denote the composite product and the primary brandin the primary category, respectively, and let sb denote thesecondary brand in the secondary category.

Before the launch of the composite product (t < t0), the util-ity of choosing a particular brand in the laundry detergent(fabric softener) is given by:

(1)

(2)

where uPhjt is the utility of buying brand j in the primarycategory at time t by household h, bPhj is the brand dummyfor brand j, and XPhjtβPh captures the effect of marketingmixes. Equation 2 represents the corresponding utility ofbuying brand k in the secondary category. However, owingto the introduction of the composite product (t > t0), wewould expect two changes. First, the number of alternativesincreases in the primary category (detergent). Second, brandexperience effects occur (from the primary/secondary brandto the composite product) as well as reciprocal effects (fromthe composite product to the primary/secondary brand). Forthe composite product, the primary brand, and the remainingbrands in the primary category, we specify the utility func-tions below.

Utility of Composite Product

(3)

First, we assume the utility of the composite product in theabove equation depends upon the brand effect bPh,cp andmarketing mix effect XPh,cp,tβPh. In addition, we introduceαPLh, αPFh, and αSLh, αSFh to represent the segment-specificbrand-experience effects, where L denotes loyal users and Fdenotes familiar users. Therefore, PL represents the loyalusers of the primary brand, SL means the loyal users of thesecondary brand, and so on. The terms αPFh and αSFh cap-ture the impact of a customer’s previous familiarity with theprimary brand and the secondary brand, respectively, on thevalue of the composite product, when the customer is afamiliar user. On the other hand, αPLh and αSLh represent thebrand experience effect on the composite product when thecustomer is a loyal user of the primary or secondary brand,respectively. Since we normalize the non-user as the base-line, we implicitly assume no brand experience effect occursif the customer has no prior experience with the primary orthe secondary brand. Finally, we also model the potentialinteraction effects for those customers who are loyal to boththe primary and the secondary brands. Let BL denote theloyal users of both the primary brand and the secondarybrand. Therefore, αBLh captures the interaction effects ofbeing loyal users of both categories.

Utility of Primary Brand

(4)

where CPE denotes a household’s experience with the com-posite product. Following the prior definition of “reciprocaleffect indicator” (Swaminathan, Fox, and Reddy 2001), weset CPE to 1 if the household has tried the composite productat time t, and 0 otherwise. Therefore, λPh captures the recip-rocal effect of adopting the composite product on the pri-mary brand. We expect λPh to be positive and significant if areciprocal effect occurs as we hypothesize.

The utility of the remaining brands in the primary categorystay the same (equation 1) as before the composite product

u b X j P t t, ,hjtP

hjP

hjtP

hP

hjtP

0β ε= + + ∀ ∈ <

u b X k S t t, ,hktS

hkS

hktS

hS

hktS

0β ε= + + ∀ ∈ <

u b I PBE 1h cp tP

h cpP

hPL

ht, , , 0α { }= + ==

PBE I PBE 0,1hPF

ht ht0 0α { }( )+ ∈

I PBE I SBE1 1hBL

ht ht0 0α { } { }+ == ==

X t t,h cp tP

hP

h cp tP

, , , , 0β ε+ + ≥

u b CPE X t t,h pb tP

h pbP

hP

ht h pb tP

hP

h pb tP

, , , , , , , 0λ β ε= + + + ≥

I SBE 1hSL

ht0α { }+ ==

SBE I SBE 0,1hSF

ht ht0 0α { }( )+ ∈

6We aggregated SKUs’ marketing mixes to brand level using marketshare as weights.


introduction. We assume the composite product introductionhas no direct effect on the utility of other brands. However,the choice probability would change owing to the shift in theutilities of the primary brand and the composite product.

For the secondary category, the number of alternativesremains the same. We only need to incorporate the reciprocaleffect into the utility of the secondary brand. Utility of thesecondary brand and the remaining brands in the secondarycategory are specified as follows.

Utility of Secondary Brand

(5)

The interpretation of the above equation is similar to that ofequation (4). The utility of remaining brands remains thesame as in equation (2).

We assume the error terms follow a multivariate normal dis-tribution with mean 0 and covariance matrix ∑. This yieldsa multinomial probit (MNP) model (Chintagunta 2002).7The probability of household h choosing brand j in week t incategory C is:

(6)

where . The calculation of the probability requires integration over the distribution ofthe error terms, and we use a GHK simulator (Hajivassiliouand Ruud 1994) to approximate this probability. We can fur-ther write down the likelihood of all the purchase outcomesin both the primary and the secondary categories of a house-hold h as:

(7)

where yPhjt (yShjt) is a purchase indicator variable that is setto 1 if brand j in the primary (secondary) category is pur-chased at trip t, and 0 otherwise.

Incorporating Heterogeneity

Significant unobserved heterogeneity across households inbrand preferences and responses to marketing variables iswell documented (Gupta and Chintagunta 1994; Kamakuraand Russell 1989). We incorporate heterogeneity with a ran-dom coefficient approach (e.g., Chintagunta, Jain, and Vil-cassim 1991; Gonul and Srinivasan 1993). Specifically, we

assume that the brand value, marketing mix coefficients, andbrand experience and reciprocal effect coefficients follow amultivariate normal distribution across households andacross both categories. The mean and the variance of thenormal distributions are estimated from the data. Let bPh bethe vertical stack of all brand value parameters in the pri-mary category, bPh = (bPh1 , bPh2 , …, bPhJ ), and let bShdenote the same brand value vector in the secondary cate-gory. Let θh be the vertical stack of all brand experience andspillover effects in both categories, θh = (αPLh, αPFh, αSLh,αSFh, αBLh, λPh, λSh). We assume: θh ~ N(θ0, Σθ), bPh ~N(bP0, ΣbP), bSh ~ N(bS0, ΣbS), βPh ~ N(βP0, ΣβP), βSh ~N(βS0, ΣβS)8. The joint log likelihood of both categories canbe computed by integrating over the heterogeneity distribu-tion. Let Θh = {bPh, bSh, βPh, βSh, θh}, and let f(Θ) denotethe joint distribution of all parameters in Θh. The log likeli-hood can be written as

We use simulated maximum likelihood to estimate the abovemodel (see Hajivassiliou and Ruud 1994 for properties ofsimulated likelihood estimator). The integration can beapproximated by simulation as follows:

(8)

where NS is the total number of simulated draws and ΘNShare random draws from the distribution f(Θ). We can thenobtain estimates by maximizing the above equation.

Estimation ResultsWe describe below the estimation results of our proposedmodel for each of the two product category pairs and thendiscuss the implications. We estimated the full model and abaseline model assuming no brand experience and reciprocaleffects. The full model outperforms the baseline modelaccording to the Bayesian Information Criterion BIC (69643vs. 70198 for detergent and softeners; and 37300 vs. 37412for automatic and manual dishwashing detergents).

Results for Laundry Detergent and Softener Categories

Table 3 lists results of laundry detergent and softener. Thecoefficient for price is negative and significant for both cate-gories, suggesting that as price increases, consumer choicelikelihood goes down. In addition, feature, display, andcoupon availability have a positive and significant effect onconsumer choice, as expected. State dependence coefficientis positive and significant, indicating a strong inertia inbrand choice. The effect of advertisement spending is sig-nificant for laundry detergent but is insignificant for fabric

u b CPE X t t,h sb tS

h sbS

hS

ht h sb tS

hS

h sb tS

, , , , , , , 0λ β ε= + + + ≥

u u k jPr Prob ,hjtC

hjtC

hktC

htC

BhtC

hjt

∫∫∫ φ ε( ) ( )= > ∀ ≠ =ε ∈

L Pr Prh hjtP y

j

J

t

T

hktS y

k

K

t

T

11 11

hjtP

hktSh

HhI

∏∏ ∏∏= ( ) ( )== ==

LLNS

Llog 1h h

ns

ns

NS

h

H

11∑∑ ( )≈ Θ

==

B u u k j s.t. ,hjt htC

hjtC

hktCε{ }= > ∀ ≠

LL L f dlog .h h h hh

H

1∫∑ ( )( ) ( )= Θ Θ Θ

=

7Households’ past purchases could influence future decisions. Statedependence captures this effect and is coded as 1 if this brand isbought at the last purchase occasion and 0 otherwise.

8The multinomial probit model should also alleviate any concernsabout the IIA assumption, as it is not likely to hold owing to the simi-larity of the composite product with the primary and secondarybrands.


softener. The brand intercepts represent the relative prefer-ence in choice with respect to a base brand (which weselected as Arm & Hammer in both categories) after control-ling for various marketing mix variables. The coefficient ofthe number of SKUs in a category is positive and significantfor the laundry detergent category, implying that consumerstend to choose brands with more SKUs available. However,this is not applicable to the fabric softener category.

For the utility of the composite product, we find that thecoefficients of the primary brand loyal user (0.57, p = .34)and the secondary brand loyal user (-.31, p = .72) are insig-nificant. These two coefficients represent the brand-experi-ence effects of a loyal user of the primary brand and the sec-ondary brand, respectively, on the value of the compositeproduct. Recall that these segment-specific results are withrespect to the baseline segment of non-users of the primaryand secondary brands. Therefore, we find no difference inthe adoption likelihood of composite products between loyalusers and non-users of the primary as well as the secondarybrands. The coefficient (2.17, p < .01) of primary brandexperience (PBE) is positive and significant, while the coef-ficient (-.08, p = .91) of secondary brand experience (SBE)is insignificant. These two coefficients capture the impact ofa customer’s previous experience or familiarity with the pri-mary and secondary brand, respectively, on the value of thecomposite product, when the customer is a familiar user.Thus, the adoption likelihood of the composite product isgreater among familiar users than among non-users of theprimary brand, but not of the secondary brand. The coeffi-cient (-2.19, p < .10) of the interaction between the primarybrand loyal user and the secondary brand loyal user is nega-tive and marginally significant, showing that adoption like-lihood of the composite product among loyal users of boththe primary and the secondary brands is lower compared tothat of either the primary brand or the secondary brand. Thecoefficient (3.15, p <.01) of the reciprocal effect of the com-posite product on the primary brand is positive and signifi-cant, showing that adoption of the composite product resultsin a positive reciprocal effect on the primary brand. We alsofind that the coefficient (1.70, p = .16) of the reciprocaleffect of the composite product on the secondary brand ispositive but not significant, which suggests no reciprocaleffect occurs with respect to the secondary brand. This find-ing is consistent with our expectation that the reciprocaleffect on the secondary brand should be either non-existentor weaker than that for the primary brand.

ConclusionsDrawing on scanner panel data, the current research foundthat relative to the low likelihood that non-users of the pri-mary brand would choose the composite product, familiarusers of the primary brand were more likely to adopt the

composite product but loyal users of the primary brand werenot. In contrast, the buying likelihood of the composite prod-uct was not different between non-users, familiar users, andloyal users of the secondary brand.9 Moreover, loyal users ofboth the primary and secondary brands have lower likeli-hood of choosing the composite product compared to non-users of either of the two partner brands. Regarding otherfindings, the choice of the composite product resulted in apositive reciprocal effect for the primary brand but did notgenerate either a positive or negative reciprocal effect for thesecondary brand and did not cannibalize host brand sales.We believe that these results make a stronger contributionthan those related to adoption likelihood of composite prod-ucts. In addition, the introduction of the composite producthelped both partner brands in several ways, includingimproving their competitive clouts, but in terms of marketshare gain, it helped the primary brand more than it helpedthe secondary brand. Finally, results of the supplementaryanalysis revealed that the introduction of the compositeproduct affected the close competitors of both the primaryand secondary brands in their respective categories, althoughthe effect was not uniform. The findings were consistentacross both composite products (Tide and Downy; Cascadeand Dawn), attesting to the potential generalizability of ourfindings. Next, we discuss implications of our findings forbrand managers, retailers, and theory.

Theoretical Implications

Our findings extend the ingredient branding literature, whichhas shown that positive equity of partner brands influencesconsumers’ stated attitude toward composite product(Simonin and Ruth 1998). However, as our findings reveal,this does not necessarily influence consumers’ revealedchoice behaviors. For example, primary brand loyal con-sumers, secondary brand loyal consumers, and secondarybrand familiar users, despite holding a positive attitudetoward the primary and secondary brands, have very lowlikelihood of adopting the complements-based compositeproduct. Thus, while brand equity and good fit between part-ner brands and categories might be adequate to influencecomposite brand attitude, among these segments other fac-tors seem to override these influences on buying the com-posite product. As we don’t have access to thought processesof consumers in these segments, we speculate on these fac-tors on the basis of our theoretical arguments. For example,primary brand loyal consumers are less likely to buy com-plements-based composite products because their exclusiveuse of the primary brand might make them believe thatadding the secondary brand to the primary brand will inter-fere with the core functionality of the product (Kirmani,

9We do not assume a full covariance matrix over all the parametersdue to computation consideration.


Table 3. Random Coefficient Probit Model Results for Laundry Detergent and Softener Categories Coefficient Coefficient Sigma Sigma Mean Std Error Mean Std ErrorPB Loyal User .57 .60 .11 .48SB Loyal User -.31 .87 .31 .30PB Loyal X SB Loyal -2.19* 1.24 1.23 1.72PBE 2.17*** .44 1.09 .45SBE -0.08 .72 .64 .60Reciprocal effect of composite product

on primary brand (CP2PB) 3.15*** 1.14 3.09 2.79Reciprocal effect of composite product

on secondary brand (CP2SB) 1.70 1.20 6.61 1.87Laundry Detergent

Arm & Hammer Base Composite product 10.48*** .43 .10 .16All 3.42*** .10 1.21 .09Cheer 9.95*** .18 1.66 .15Era 3.73*** .13 1.6 .09Private Label -4.87*** .15 1.7 .13Purex 1.44*** .09 1.74 .17Tide 6.21*** .21 2.14 .17Wisk 6.22*** .12 .12 .16Xtra -6.36*** .15 .96 .12Price -3.99*** .05 1.13 .03Feature 2.34*** .08 1.48 .09Display .94*** .07 .56 .08Coupon availability 1.82*** .27 2.03 .41State dependence 12.55*** .12 .15 .17Ln (advt. spending) .05*** .01 .05 .01UPC count .28*** .01 .04 .01

Fabric SoftenerArm & Hammer Base All -.99*** .21 .80 .21Bounce -.96*** .32 .98 .30Cling Free -.46 .30 .57 .27Downy 3.76*** .49 2.07 .17Private Label -1.64*** .33 2.57 .14Snuggle 1.26*** .36 1.49 .14Price -1.00*** .03 .71 .03Feature 3.54*** .11 1.95 .14Display .86*** .11 .29 .19Coupon availability 1.49*** .25 1.86 .73State dependence 12.75*** .21 1.74 .17Ln (advt. spending) -.00 .01 .02 .02UPC count .00 .04 .01 .01

Note: 1. We also controlled for new product introductions by incorporating dummy variables in both laundry detergent and softener categories. 2. *** p < .01; ** p < .05; * p < .1


Sood, and Bridges 1999; Roedder-John, Loken, and Joiner1998). Similarly, familiar users of the secondary brandmight be less willing to buy the composite product becauseto adopt it, they would have to give up their current brand inthe primary category.

Our research has a few limitations. While strengtheninginternal validity, the extensive criteria used in selectingcomposite products limited us to two replicates. Theseselection criteria severely restricted the number of compos-ite product replicates we could work with, but they allowedus to be more confident in attributing our findings to positedtheoretical arguments (e.g., attributing choice of the com-posite product to users familiar with the primary and sec-ondary brands or attributing reciprocal effect on choices ofprimary and secondary brands). Future research can relaxthese criteria to see whether our findings extend to morecomposite products. Moreover, our research used con-sumers’ choices to test the hypotheses, but without access toconsumers’ thought process, we can only speculate abouttheoretical reasons underlining our findings. Another limi-tation is that the complements-based composite productsselected were “same” use situation products—that is, nei-ther partner brand could be used outside of the common use

situation of washing clothes or cleaning dishes. An interest-ing investigation would explore how our findings wouldchange if this assumption of common use is relaxed byinvestigating composite products such as the El Fudgesandwich cookie with Nestle’s Butter Finger candy, inwhich consumers use or consume the primary and second-ary brands in distinct situations. Since we have used datafrom two sets of complementary categories belonging to asingle chain located in a particular market area, futureresearch could use data from multiple stores, categories,and market areas to generalize our findings and attribute theinfluence of any of these characteristics on consumerchoice. An interesting avenue of research would be thestudy of the moderating influence of various factors, suchas the quality of the composite product or the perceivedvalue, on composite product choice. We also acknowledgethat this study lacks information about manufacturing costs,which could help investigate the impact of composite prod-ucts on profitability of various brands (George, Kumar, andGrewal 2013). Finally, future research can use multi- andcross-category models (Russell et al. 1999; Pancras, Gauriand Talukdar 2013) to model both primary and secondarycategories to yield more insights into the substitution pat-terns induced by composite products.

Figure 1 Conceptual Framework

ox iMng itkeraMndsarBlaocF

of x iMng itkeraMarBng itpeomC

ofx ndsa

C

etiposomCtoducrP

echoiC

g p

ndarBy ramirP

enceirxpeE

narBy rondaceS

laocpriceRstcefeEff

nd

enceirxpeE


ReferencesChintagunta, Pradeep K. (2002), “Investigating Category

Pricing Behavior at a Retail Chain,” Journal of Market-ing Research, 39(2), 141-154.

_______, Dipak C. Jain, and Naufel J. Vilcassim (1991),“Investigating Heterogeneity in Brand Preferences inLogit Models for Panel Data,” Journal of MarketingResearch, 28(4), 417-428.

Desai, Kalpesh Kaushik and Kevin Lane Keller (2002),“The Effects of Ingredient Branding Strategies on HostBrand Extendibility,” Journal of Marketing, 66(1), 73-93.

Elrod, T., G. J. Russell, A. D. Shocker, R. L. Andrews, L.Bacon, B. L. Bayus, J.D. Carroll, R. M. Johnson, W. A.Kamakura, P. Lenk, J. A. Mazanec, V.R. Rao and V.Shankar (2002), “Inferring market structure from cus-tomer response to competing and complementary prod-ucts,” Marketing Letters, 13(3), 221-232.

Erdem, Tulin (1998), “An Empirical Analysis of UmbrellaBranding,” Journal of Marketing Research, 35(3), 339-351

Fader, Peter S. and Bruce G.S. Hardie (1996), “ModelingConsumer Choice among SKUs,” Journal of MarketingResearch, 33 (November), 442–452.

George, Morris, V. Kumar and Dhruv Grewal (2013),“Maximizing Profits for a Multi-Category CatalogRetailer,” Journal of Retailing, 89(4), 374-396.

Geylani Tansev, J. Jeffrey Inman and Frenkel T. Hofstede(2008), “Image Reinforcement or Impairment: TheEffects of Co-Branding on Attribute Uncertainty”, Mar-keting Science, 27(4), 730-744.

Gonul, F. and K. Srinivasan (1993), “Modeling MultipleSources of Heterogeneity in Multinomial Logit Models:Methodological and Managerial Issues,” Marketing Sci-ence, 12(3), 213-229.

Guadagni, P. and J.D.C. Little (1983), “A Logit Model ofBrand Choice Calibrated on Scanner Data,” MarketingScience, 2(2), 203-238.

Gupta, Sachin and Pradeep K. Chintagunta (1994), “OnUsing Demographic Variables to Determine SegmentMembership in Logit Mixture Models,” Journal of Mar-keting Research, 31(1), 128-136.

Hajivassiliou, V. and P. Ruud (1994), “Classical EstimationMethods for LDV Models using Simulation,” Handbookof Econometrics, Vol. IV, R. Engle and D. McFadden eds.Amsterdam: Elsevier Science Publishers B.V.

Kamakura, Wagner and Gary J. Russell (1989), “A Proba-bilistic Choice Model for Market Segmentation and Elas-

ticity Structure”, Journal of Marketing Research, 26(4),379-390.

Keller, Kevin L. and David A. Aaker (1992), “The Effects ofSequential Introductions of

Brand Extensions,” Journal of Marketing Research, 29(1),35-50.

___________, and Donald R. Lehmann (2006), “Brands andBranding: Research Findings and Future Priorities,” Mar-keting Science, 25 (6), 740-59.

Kirmani, Amna, Sanjay Sood, and Sheri Bridges (1999),“The Ownership Effect in Consumer Responses to BrandLine Stretches,” Journal of Marketing, 63 (January), 88-101.

Krishnamurthi, L. and S. P. Raj (1988), “A Model of BrandChoice and Purchase Quantity Price Sensitivities,” Mar-keting Science, 7(1), 1-20.

Lattin, J. M. and R. E. Bucklin (1989), “Reference Effects ofPrice and Promotion on Brand Choice Behavior,” Journalof Marketing Research, 26 (3), 299-310.

Mehta, Nitin (2007), “Investigating Consumers’ PurchaseIncidence and Brand Choice Decisions across MultipleProduct Categories: A Theoretical and Empirical Analy-sis,” Marketing Science, 26(2), 196-217.

Pancras Joseph, Dinesh K. Gauri and Debabrata Talukdar(2013), “Loss Leaders and Cross-Category Retailer Pass-through: A Bayesian Multilevel Analysis,” Journal ofRetailing, 89(2), 140-157.

Park, C.W., S. Youl Jun and A.D. Shocker (1996), “Compos-ite Branding Alliances: An Investigation of Extension andFeedback Effects,” Journal of Marketing Research,33(4), 453-466.

Roedder-John, Deborah, Barbara Loken, and ChristopherJoiner (1998), ‘The Negative Impact of Extensions: CanFlagship Products Be Diluted?” Journal of Marketing, 62(1), 19-32.

Russell, G. J., S. Ratneshwar,, A. D. Shocker, D. Bell, A.Bodapati, A. Degeratu, L. Hildebrandt, N. Kim, S.Ramaswami and V. H. Shankar (1999), “Multiple-Cate-gory Decision-Making: Review and Synthesis,” Market-ing Letters, 10(3), 319-332.

Simonin, B.L. and J.A. Ruth (1998), “Is a Company Knownby the Company it Keeps? Assessing the SpilloverEffects of Brand Alliances on Consumer Brand Atti-tudes.” Journal of Marketing Research, 35(1), 30-42.

Swaminathan, Vanitha, Richard J. Fox, and Srinivas K.Reddy (2001), “The Impact of Brand Extension Introduc-tion on Choice,” Journal of Marketing, 65 (October), 1-15.

Superstars in general seem to dominate many industries.Superstar individuals or products “possess unique and supe-rior attributes or skills” and command a disproportionatelylarge payoff (Binken and Stremersch 2009; Rosen 1981).Examples of such superstars abound in most industries—Tiger Woods in golf, Steven Levitt among economists,Microsoft Word in word processing, IBM in consulting. Inthe entertainment industry especially, superstars appear todominate product segments—Grand Theft Auto in videogames, Da Vinci Code in fiction, Brad Pitt among actors,Game of Thrones on cable television, Howard Stern on radio,Angry Birds among mobile phone apps, Harry Potter amongmovies, etc. Binken and Stremersch (2009), in a seminal arti-cle in the marketing literature, discuss the importance ofsuperstar video games in the context of the video game indus-try and show that the introduction of game superstars affectsthe sales of hardware (console) significantly positively.

Our data set is provided by the NPD Group, and includesmonthly point-of-sale data on prices and quantities for videogame consoles from January 1995 through October 2007from roughly 65% of US retailers. In total, the data set con-tains average price and quantity data for 15 consoles over 4distinct product generations.

Using this dataset, our paper extends the results of Binken andStremersch (2009) in several ways. First, we build on theirresults and show that the presence of superstar software posi-tively and significantly affects console market share. Second,and importantly, we show that such superstar effects vary dra-matically over the product generation’s (console) lifecycle.Specifically we find that superstars are significant for consolemarket share early on, especially in the Introductory and

Growth phases of the product generation lifecycle but notimportant later on. This finding provides interesting manage-rial insights which we discuss later and serves as key depar-tures from extant results. Third, we find that regular games(non-superstars) are important later on especially in theDecline phase of the generation product lifecycle but notimportant earlier. Thus managerial strategies regarding super-stars should vary with the console generation lifecycle.Fourth, Binken and Stremersch (2009) do not consider asimultaneous equation system where both hardware and soft-ware impact each other (feedback loop). However, followingStremersch et al. (2007), we jointly estimate both the demandfor hardware and the supply of regular and superstar softwarewhich make our results more robust. Including this complex-ity (feedback loop) has the benefit of informing managerswhen they can leverage their installed base to attract super-stars to their console. Our results show that the effect of theinstalled base on the supply of superstars is significant in theGrowth phase and not in the Introduction phase. Also, theinstalled base has a positive and significant effect on superstarsupply in the Decline phase—when the effect of superstars onconsole demand is weakest. This interesting result and insightregarding the criticality of the network effect on the supply ofsuperstars was ignored in prior research and would not be pos-sible without considering the feedback loop that we model.Fifth, software and hardware sale are usually affected byomitted variables that pose endogeneity problems, an issuenot addressed by Binken and Stremersch (2009). In this paper,we control for such endogeneity by using several reasonableinstruments with instrumental variables regression techniquesvia generalized method of moments (GMM) estimations.

References are available on request.

The Effect of Superstar Software in theVideo Game Industry: The ModeratingRole of Product Generation Lifecycles Richard Gretz, Bradley UniversitySuman Basuroy, University of Texas at San Antonio

Keywords: indirect network effect, superstars, software, hardware, new product introductions, product generation lifecycle, video games

EXTENDED ABSTRACT

For further information contact: Richard Gretz, Associate Professor, Bradley University (e-mail: [email protected]).


Research Question Marketing managers and individuals making a variety ofdecisions (decision makers hereinafter) are likely to con-tinue a failing course of action and escalate commitment ifthey are responsible for initiating the course (Boulding,Morgan, and Staelin 1997). However, none of the research todate shows that responsibility to initiate a course of action(here after personal agency) mitigates (i.e. decrease) escala-tion of commitment conditionally. Moreover, debate con-cerning the process to explain this phenomenon is thriving(Sleesman et al., 2012).

In this paper, I integrate the research on escalation of com-mitment and the research on regret, a negative emotion thatis experienced based on cognitive evaluations of an adverseoutcome (Zeelenberg, Van Dijk, and Manstead 2000a), toexplicate the conditional effects of personal agency on thephenomenon. I propose that likelihood of receiving a desiredfinal outcome (here after outcome reversibility) jointly withthe availability of the information about superior foregoneoutcome moderates the effect of personal agency on escala-tion of commitment. I also hypothesize that regret mediatesthe effect of personal agency on escalation of commitment topropose regret regulation (Zeelenberg and Pieters 2007) as atheoretical rationale for the phenomenon.

Method and Data I recruited 192 business students (mean age 21 years, male49%, experience 4.25 years) and evaluated their escalationof commitment decision on a stock investment using a 2(personal agency: self select vs. assigned) x 2 (probability ofreceiving a desired outcome subsequent to interim loss fromthe stock investment: 65% chance versus 25% chance)between-subjects design. At the outset of the study, Imanipulated personal agency by asking half of the partici-pants to select and justify the stock they would choose to

receive as a gift (Bobocel and Meyer 1994). The other halfof the participants had no choice.

Subsequently, I manipulated reversibility of the adverse out-come by reporting an interim loss from the investment and aprobability (65% versus 25%) that the future outcome of thestock will deliver desired outcome. I also reported the supe-rior performance of the foregone stock investment to all par-ticipants. Then, I measured each participant’s regret, self-attribution for the loss and perception of the superiority of theforegone option. In addition, as measures of attitudinal andbehavioral escalation of commitment, I assessed each partici-pant’s intentions to sell the stock and the actual amount of thestock they sold. Furthermore, I gave them bonus cash andmeasured their additional investment in the stock.

Summary of Findings Study results suggest that personal agency mitigates per-ceived superiority of the foregone option when thereversibility of the received outcome is high, and amplifiesit when the reversibility of the received outcome is low. Onthe other hand, results suggest that personal agency not onlytriggers self-attribution when the outcome of the foregoneoption is superior, but it also further amplifies self-attribu-tion when the reversibility of the received outcome isreduced. More importantly, results show that when the out-come of the foregone option is superior, the effect of per-sonal agency on regret and escalation of commitment is con-tingent upon the reversibility of the received outcome. Onthe one hand, when the reversibility of the received outcomeis high, participants who self-choose a course of actionexperience less regret, which, in turn, increases their escala-tion of commitment. On the other hand, when the reversibil-ity of the received outcome low, those who self-chooseexperience more regret, which, in turn, decreases their esca-lation of commitment. Accordingly, a mediation test sug-

The Mitigating Effect of Personal Agencyon Escalation of CommitmentSunil Contractor, Johns Hopkins University

Keywords: escalation of commitment, regret, decision-making, foregone outcome, personal responsibility

EXTENDED ABSTRACT

For further information contact: Sunil Contractor, Assistant Professor of Marketing, Johns Hopkins University (e-mail: [email protected]).



gests that regret mediates the conditional effects of personalagency on escalation of commitment.

Key Contributions Study results show that personal agency decreases orincreases escalation of commitment when the outcome ofthe foregone option is superior and the reversibility of thereceived adverse outcome is low or high, respectively.Therefore, this is the first study to demonstrate the mitigat-ing effect of personal agency on escalation of commitment.The results also suggest that the knowledge of the foregoneoption’s outcome and the reversibility of the received out-come are important characteristics of an escalation situation.While proposing self-justification theory as a rational forescalation of commitment, Staw (1976) asserts that indi-

viduals attempt to avoid self-attribution of causality whentheir behavior leads to negative consequences or results inpersonal failure. However, I find that those who are respon-sible for initiating a failing course of action not only engagein self-attribution for the adverse outcome, but they amplifytheir self-attribution when the probability of recovering froma loss decreases so that the negative consequence or the per-sonal failure is more apparent. Moreover, while researchershave proposed several other theories to explain the amplify-ing effect of personal agency on escalation of commitment,none explains the mitigating effect. Therefore, my researchsuggests regret regulation as a theoretical rationale for theeffect of personal agency on escalation.


Research QuestionThe literature in marketing, management, strategy and inno-vation has evolved in two important directions. One,antecedents to innovation have been studied at the organi-zation or project level. There has not been much work on anorganization-level innovation capability and the work thathas been done is largely conceptual. Two, there has been alot written about two kinds of innovation, incremental andradical innovation. However not much attention has beenpaid to the issue of whether the antecedents and outcomesof radical product innovation are different than those forincremental innovation. It is widely recognized that compa-nies like Amazon, Apple, P & G, and Google are better atbeing pioneers and launching radical innovations whereasDell, Microsoft and Samsung are arguably better at beingfollowers or launching incremental innovations. This papertakes a focused look at the antecedents of an organizationalcapability called radical product innovation capability andexplores how that results in specific consequences for newproduct launches such as being first to market and productperformance.

Method and DataWe developed an online survey with multiple-item measuresfor the constructs of interest. The items were presented inscrambled order so that underlying constructs could not beidentified. Some items had the wording reversed at periodicintervals. The survey questionnaire was pre-tested with 9executives familiar with innovation processes in their com-pany and revised for clarity. A list of 6000 email addresses

was purchased from a vendor, of senior managers in US com-panies across many industries. An email solicitation with thelink to the online survey was sent to each person addressed byname. They were requested to answer the survey if they werefamiliar with a major new product or service their companyhad launched within the past 3 years. Respondents were askedto answer the questionnaire with that product in mind. Theywere also asked to voluntarily provide the name of the productor a web page address where one could obtain more informa-tion about the product (81% of respondents provided this). Wereceived 97 responses of which 95 were usable. The mainindustries in our sample were Computer and Electronic Prod-uct Manufacturing (34%), Publishing but not Internet (27%),Machinery Manufacturing (13%), and Professional, Scien-tific, and Technical Services (12%).

Summary of FindingsWe have a 3-stage model and do path analysis results usingOLS regression. We use two control variables at every stage,one for the environment (market uncertainty) and one for theage of the business. In Stage 1, we find that all three of ourantecedents (risk tolerance, learning to explore new compe-tencies, innovation process) are positively related to RPI capa-bility. In Stage 2 we find that RPI Capability is positivelyrelated to being First to Market. And in Stage 3 we find thatFirst to Market is not related to RPI Performance, however,RPI Capability is positively related to RPI Performance. Wetest the moderating effect of being First to Market on the posi-tive relationship between RPI Capability and RPI Perform-ance and find there is a positive moderating effect.

An Exploratory Study of Antecedents andConsequences of Radical Product Innovation CapabilitySanjit Sengupta, San Francisco State University Stanley F. Slater, Colorado State University Jakki J. Mohr, University of Montana

Keywords: innovation, capability, first-to-market, performance

EXTENDED ABSTRACT

For further information contact: Sanjit Sengupta, Professor of Marketing, San Francisco State University (e-mail: [email protected]).



Key ContributionsThis paper makes several contributions to the literature. First,it conceptualizes measures and empirically tests the conceptof radical product innovation capability, as the central con-struct, which enables firms to consistently bring out new tothe world products every few years. Second, it providesempirical support for the antecedents to radical product inno-vation. In order to develop, nurture and maintain this capa-bility firms need to take risks and tolerate failure as part oftheir culture of innovation. They need to encourage employ-ees to explore and learn about new competencies (technolo-gies, products, markets) and they also need to have internalprocesses in place to incubate radical ideas and convert theminto big business opportunities. Third, we show that a radicalproduct innovation capability does result in pioneering orfirst-mover advantage. Consistent with equivocal findings in

the literature, (Tellis and Golder 1996) we find no direct rela-tionship between being first to market and product perform-ance. However, if a firm has RPI Capability and is first to mar-ket, we do find a positive impact on patent performance.

There are some interesting managerial implications of ourfindings. What separates Amazon, Apple and Google fromDell, Microsoft and Samsung, we believe, is the first threecompanies have had radical product innovation capabilitythrough most of their history whereas the latter three havenot. Just being first to market doesn’t result in successfulproduct performance. Being first to market together withradical product innovation capability is needed to enhanceproduct performance.



Research QuestionLow-cost innovations, defined as new products or servicesthat organizations direct at consumers who have a lowerwillingness-to-pay, enable firm growth and help drivesocieties’ prosperity. However, research has generated aplethora of different theories and frameworks to investigatelow-cost innovation. Consequently, this study seeks answersto the following questions:

1) Which theories relate to low-cost innovation?2) How do low-cost innovations create value for consumers?3) How can companies capture the value they have created?

Method and Data A systematic review was conducted to provide collectiveinsights through theoretical synthesis into fields and sub-fields following a structured, reproducible methodology.Using the different sub-fields of low-cost innovation asinputs into the database search and systematically eliminat-ing irrelevant papers results in a sample of 97 relevant, peer-reviewed articles. Starting from a basic structure of valuecreation and value capture, I used the text analysis softwareMaxQDA to code each of the 97 articles. Finally, a referencenetwork analysis reveals clusters in low-cost innovationresearch.

Summary of Findings The results of the network analysis imply contingencies con-cerning the introducer (i.e., incumbents vs. entrants) and thetarget market (i.e., emerging vs. developed market). Apply-ing the contingency lens in combination with the value crea-tion and value capture perspective in the content analysis

leads to a set of consistent factors (i.e., valid across settings)and contextual factors (i.e., valid in particular settings). Con-sistent value creation factors for low-cost innovation includea lower price, a focus on key performance dimensions andnew value combinations. Contextual value creation mecha-nisms in emerging markets comprise innovating financingmodels and creating access. Firms across different settingsare more likely to capture value from low-cost innovationwhen they can use economies of scale and when they pos-sess a low-cost innovation culture. In addition, incumbentsbenefit from creating an ambidextrous organization andfirms in emerging markets need a business model orientationas well as unique low-cost innovation processes.

Key Contributions Researchers use a multitude of different theoretical lenses toanalyze products and services that organizations direct atconsumers who have a lower WTP than the going rate. Byhighlighting overlapping theories from disruptive innova-tion, emerging markets, strategy and entrepreneurship, anddefining low-cost innovation from a consumer point of view,this study provides a basis for further research on this issue.The systematic review reveals consistent factors for bothvalue creation and value capture that are valid across differ-ent settings and firm types. In addition, the analysis revealedunique factors that are contingent on the setting and firmtype. Consequently, low-cost innovation theory can beunderstood as a nested model. The proposed model offersthe opportunity to combine research streams on a globallevel as well as on a context specific level.


Avoiding a Babylonian Confusion: A Systematic Review on Low-Cost InnovationRonny Reinhardt, Technische Universität Dresden

Keywords: low-cost innovation, disruptive innovation, BOP innovation, frugal innovation, systematic review

EXTENDED ABSTRACT

For further information contact: Ronny Reinhardt, Department of Business and Economics, Chair for Entrepreneurship and Innovation, Technische Univer-sität Dresden (e-mail: [email protected]).


Research QuestionsBased on an extensive literature review we find that there isa lack of research (1) addressing the effectiveness of collab-oration breadth for the development of successful really newproducts (RNPs) and (2) investigating the impact of collab-oration breadth in the different stages of the new productdevelopment (NPD) process. Collaboration breadth refers tothe number of different types of external collaboration part-ners (Laursen and Salter 2013). The deficiency of research issurprising because the various stages of the NPD process,i.e. the idea generation, research and development, design,test and evaluation, and market introduction stage, have dif-ferent knowledge requirements and thus, lead to differentlevels of uncertainty. These knowledge requirements can bepotentially addressed by various and different external col-laboration partners. Thus, different levels of collaborationbreadth can be an effective way to integrate the requiredknowledge in the NPD process and to reduce stage-specificuncertainties. Therefore, the main research questions of thepaper are: (1) how does collaboration breadth impact theperformance of RNPs overall and (2) how does collabora-tion breadth impact the performance of RNPs specifically inthe various stages of the NPD process? (3) whom to collab-orate with in which stage of the NPD process?

Method and DataWe used the Mannheim Innovation Panel from the Centerfor European Economic Research as our data source. The

yearly conducted large-scale survey designed as a panel rep-resents the German part of the Community Innovation Sur-vey (CIS). The definitions of constructs and the methodol-ogy of the survey follow the Oslo Manual (OECD/Eurostat2005). The CIS data have been previously used in innova-tion research (e.g., Van Beers and Zand 2014; Laursen andSalter 2006). Of the total sample of manufacturing and serv-ice firms (6,110 firms), 1,718 firms (28.1%) indicated thatthey were actively engaged in the development of RNPs,thus constituting our effective sample size.

The dependent variable ‘performance of RNPs’ is heavilyleft-censored as it measures the performance of RNPs in per-centage of sales that ranges between 0 and 100. Because ofleft-censoring we applied a Tobit regression as recom-mended in the literature (Greene 2011).

Collaboration breadth is measured as sum of six differenttypes of external collaboration partners (customers B2B,customer B2C, suppliers, service providers, competitors,and universities and research institutes) that ranges from 0 to6 (Laursen and Salter 2013). We included firm size, firmage, R&D intensity, human capital, firm location and indus-try as control variables in the analysis.

Summary of Findings Based on the relational view, the findings reveal evidence ofa general ‘over-collaboration’ that means an inverted U-

Development of Successful Really NewProducts: The “Over-Collaboration” Effectat Different Stages of the New ProductDevelopment Process Johannes S. Deker, University of Mannheim, Germany Monika C. Schuhmacher, University of Mannheim, Germany Sabine Kuester, University of Mannheim, Germany

Keywords: collaboration breadth, really new products, new product development, process stages

EXTENDED ABSTRACT

For further information contact: Johannes Deker, Chair of Marketing and Innovation, University of Mannheim (e-mail: [email protected]).


shaped relationship between collaboration breadth and theperformance of RNPs over the entire NPD process and atevery stage of the NPD process. In addition, we see acrossthe different stages of the NPD process, that the develop-ment of RNPs benefits the most from collaboration breadthin the initial idea generation stage, while the optimal numberof collaboration partners decreases towards the end of theNPD process.

Furthermore, we identify which external collaboration part-ners should be integrated in the different stages of the NPDprocess. We demonstrate that B2B customers are the mostimportant external collaboration partners as they contributeto the performance of RNPs at all stages of the NPD process.In contrast, the knowledge of suppliers seems to be very spe-cific to the development of RNPs in the R&D/constructionstage. Whereas universities and research institutes appear asimportant collaboration partners in the first stages of theNPD process, service providers gain importance in the laterstages of the NPD process.

Key Contributions First, we contribute to research on the impact of collabora-tion breadth on the performance of RNPs which has been

called for previously (e.g., Knudsen 2007). We garnerintriguing evidence for the effect of ‘over-collaboration’, anon-linear relationship between collaboration breadth andperformance of RNPs across and at each and every stage ofthe NPD process. By exploring the influence of collabora-tion breadth on the performance of RNPs in the differentstages of the NPD process, we address general calls forresearch for a stage-specific investigation of the relationshipbetween collaboration and successful NPD (e.g., Homburgand Kuehnl 2013). Second, with the focus on RNPs, we addto the understanding on how firms can improve the develop-ment and performance of the most important and criticaltype of new products represented by RNPs. Third, we pro-vide managerial implications for managers who seek to opti-mize the number of collaboration partners in the differentstages of developing RNPs as called for previously (e.g., DiBenedetto 2012). Lastly, we make a theoretical contributionto the relational view by showing that the positive impact ifcollaboration with diverse external partners in NPD has itslimits.



Research QuestionsOver the years, firms have been reinventing their innovationapproaches, anticipating changes in society and markets.Current trends suggest that businesses are witnessing theemergence of the next big innovation wave, driven by socie-tal and environmental needs. Increasingly, firms are lookingat the impact of their operations on the environment as wellas the positive or negative impact on the communities inwhich they operate. Understanding these impacts can driveimprovements in corporate strategy, in developing newproducts more creatively and competitively, and ultimatelydetermine a firm’s innovation performance. As such, thisresearch explores how the sustainability efforts of innovat-ing firms affects their new product program (NPP) perform-ance, i.e., the extent to which the new product program issuccessful and meets its performance target. This researchalso attempt to understand how firms draw on and success-fully utilize external ideas and knowledge in their new prod-uct development.

Method and DataThe data analyzed is drawn from the 2012 Product Develop-ment and Management Association (PDMA)’s ComparativePerformance Assessment Study (CPAS), which was admin-istered on a global scale using Global Park’s online surveytool. Emails containing a link to get a code that allowed eachrespondent to access the online survey were sent to PDMAmembers and PDMA contacts. A total of 1,167 codes weresent to key respondents for their business unit with instruc-tions on how to fill out the survey, yielding 453 useable sur-veys. The final sample consists of firms from North America(198), Asia (149), Europe (61), and elsewhere (45) acrossmultiple industries. Goods manufacturers make up 56% of

the sample, with technology companies comprising 45.4%and business-to-business companies accounting for 56.4%of responding companies, while companies with less than$100 million in revenue make up 63.3% of the sample. Wetest our hypotheses using multiple regressions with, if nec-essary, relevant interactions terms.

Summary of FindingsThe findings of the study suggest that sustainability canenhance firm performance through its positive impact on newproduct innovation, and customer focus mediates the innova-tion outcomes of a sustainability orientation. Furthermore, ourfindings suggest that this mediated link varies significantlyacross firms with different levels of open innovation activities.While open innovation activities aimed at gathering marketinsights enhances customer focus directly, open innovationactivities aimed at technical problem solving accentuates thelink between customer focus – NPP performance.

Key ContributionsThis study advances our knowledge about the business caseof sustainability by documenting the positive associationbetween sustainability orientation and NPP performance.Since new product innovation is critical to the profitabilityand long-term performance of any firm, our findings suggestthat sustainability can improve firm performance through itspositive impact on new product innovation. More impor-tantly, we identified a route through which sustainability ori-entation is related to NPP performance. Our results indicatethat a firm’s sustainability orientation helps enhance its cus-tomer focus in the NPP, in turn customer focus positivelyaffect the innovation outcomes. The mediating role of cus-tomer focus is important for two reasons. First, it extends the

Sustainability, Open Innovation, and NewProduct Program SuccessShuili Du, University of New HampshireLudwig Bstieler, University of New Hampshire Goksel Yalcinkaya, University of New Hampshire

Keywords: open innovation, sustainability orientation, customer focus, new product success

EXTENDED ABSTRACT

For further information contact: Goksel Yalcinkaya, University of New Hampshire (e-mail: [email protected]).


sustainability literature by uncovering a previously ignoredoutcome (i.e., customer focus) of sustainability orientation.Second, it extends the literature on customer focus byuncovering the antecedents (i.e., sustainability) of customerfocus. Furthermore, our study highlights the importance ofdifferentiating between two types of open innovation activi-ties, those aimed at gathering market insights and those

aimed at technical problem solving, as they function differ-ently to enhance NPP performance. This suggests that man-agers should leverage these two types of open innovationactivities differently in various phases of new product devel-opment program.



Research QuestionThe literature exploring co-creation, the process of inte-grating customers in the new product development process,has grown rapidly in the last decade. The extant work hasestablished co-creation as a powerful tool for both innova-tion and improving customer perceptions of the firm.While co-creation is widely used within B2B markets,there is little work exploring the use of co-creation withinB2B channels. Additionally, while co-creation is touted asa positive interaction for business, there may also be nega-tive relationship outcomes of co-creation in B2B relation-ships. This research takes a novel approach to B2B co-creation and investigates what happens when channelpartners co-create a product together and the new productperforms poorly in the marketplace. This conceptual studyseeks to answer three main questions. First, what effectmay failure of a co-created product have of B2B channelrelationships? Second, how do channel partners assessblame for the product failure and what role does blameplay in explaining negative relationship outcomes. Third,what characteristics of the relationship and what character-istics of the new product moderate the effect of productfailure on B2B partnerships? This paper takes a conceptualapproach and offers propositions regarding the effect ofnew product failure on B2B relationships.

Key ContributionsThis research makes some important theoretical and mana-gerial contributions. First, the co-creation literature in theB2B space is limited and this study sheds light on issueswithin B2B co-creation. Although new product developmentis an essential aspect of B2B success, very little is under-stood about the dynamics that exists between channel part-

ners resulting from the co-creation process. This researchproposes that the degree to which channel manufacturersassess blame to customers for the new product failureimpacts channel relationships. Second, the literature on thenegative outcomes of co-creation is also limited. Much ofthe co-creation literature discusses the positive productionand relational benefits of co-creation, but co-creation is alsobound to have negative nuances that have yet to be consid-ered. Co-creation involves new types of interactionsbetween partners which could create new avenues for fric-tion. This paper helps researchers develop a theoretical con-text for which to understand the co-creation process and howto study problems that may arise.

Summary of FindingsUsing the theoretical underpinnings taken from the new prod-uct literature and the relationship marketing literature, thisstudy proposes a conceptual model that aids future research inbetter understanding all the nuances of co-creation in B2Bsettings. Using the theory of self-serving bias, this studyshows how co-created new product failures can cause B2Brelationships to deteriorate. Relationship partners are likelyto assess blame to each other when a co-created new productperforms worse than expected. This blame causes increasedvolatility in relationship sentiments such as relationship con-tinuity and performance. This paper also proposes modera-tors that strengthen and weaken the effect of new productfailure on relationships. Specifically, this study shows howattributes of the relationship and attributes of the co-createdproduct impact how channel partners assess blame for co-created product failure.


It’s All Your Fault! Attributing Blame forCo-Created New Product Failures in B2BRelationshipsB.J. Allen, University of Texas at San Antonio

Keywords: co-creation, B2B channels, new product development, product failure

EXTENDED ABSTRACT

For further information contact: B.J. Allen, PhD student, University of Texas at San Antonio (e-mail: [email protected]).

Part F Innovation and New Products - Amazon...

Documents

Transcript of Part F Innovation and New Products - Amazon...