Co-Creating Benefits in Social Media Contests and its Effects on Purchase Behaviors

Download 41.4 Kb.
Size41.4 Kb.

Co-Creating Benefits in Social Media Contests and its Effects on Purchase Behaviors
Edward C. Malthouse, Northwestern University, USA

Mark Vandenbosch, University of Western Ontario, Canada

Su Jung Kim, Northwestern University, USA
Corresponding Author

Edward C Malthouse

Northwestern University

1870 Campus Drive

Evanston, IL 60208 USA


Co-Creating Benefits in Social Media Contests and its Effects on Purchase Behavior
This paper examines how participating in social media contests in which the consumer is asked to co-create the core brand benefit affects subsequent purchase behaviors. Using data from Canada's Air Miles Reward Program and three separate contests, we find that participation increases mile accumulation behavior over a control group and after controlling for pre-contest behavior. We propose an explanation for the effect, and discuss its implications and applications.
The Internet and social media are enabling many new forms of “advertising.” The first decade or so of Internet advertising–largely consisting of display “banner” ads and email blasts and later by search–followed the approach of traditional print, broadcast and direct advertising, where the advertiser “exposes” passive eyeballs to some message, perhaps with a “call to action” inviting a click. The rise of social media and the proliferation of mobile devices are enabling new forms of advertising that are more participatory and interactive. While in the past a firm would decide on the positioning of its brand and then use advertising to broadcast messages conveying the brand meaning, many firms are experimenting with allowing customers to co-create brand meaning, advertising messages, and other parts of the value chain (Vargo & Lusch, 2004).
One form of co-creation is contests in social media forums, where customers create “user-generated content” (UCG) in a public forum about a brand and are compensated by either a small reward or the chance of winning a larger reward. Such contests are becoming increasingly common. An older example is the Dove Cream Oil Body Wash ad contest shown during the 2007 Academy Awards (Deighton, 2008). Contestants were encouraged to create an ad for the new Dove product. A similar and more current example is the Pepsi MAX contest, in which Doritos and Pepsi rewarded winners by airing six consumer-created ads during the 2012 Super Bowl. Sony is sponsoring the “make believe Giveaway A Day” contest on their Facebook community, in which they will give away Blu-Ray Disc players and other electronics devices. Entrants must share their email address, which is a lower level of involvement than creating a TV advertisement, as in the case of Dove or Pepsi. Kit Kat Canada recently sponsored the “Game Time Give Away” on Facebook. Participants visited the Facebook page and completed an entry form. Prizes included NFL beverage pails, T-Shirts, footballs, and pens, and chocolate.
The focus here is on contests requiring co-creation. We make three contributions. First, while such contests are becoming more common, there are no published studies measuring their effect on purchase behaviors. We test whether participating in such contests affects purchase behaviors and illustrate ROI calculations. Second, we propose an explanation for the effect, which will help organizations create more effective contests going forward. Third, we will investigate the longevity of the effect—after participating in such a contest, how long is behavior affected?

Why Contests Should Work
This section proposes an explanation for why contests work, which informs their design. Contests are an effective way to engage consumers because they encourage them to share their personalized experiences with a brand, which is an essential element of value co-creation (Prahalad & Ramaswamy, 2000, 2004). Prahalad and Ramaswamy (2004) emphasized that the current shift from a firm-centric to a consumer-centric view of the market requires companies to provide unique value to customers and the base of that differentiation is the consumer's experience with the brand. Rather than being (possibly) exposed to an advertiser-created message, contests ask consumers to think about the brand and create their own message based on their past or expected future experiences with the brand. This processing should be a potent form of advertising if it is executed properly. Associations should be more salient and personal when consumers think through their relationship with brand, as opposed to being told what a brand should mean to them by an ad.
These contests can be an example of "encounter processes" (Payne, Storbacka, and Frow, 2008), which refer to "the process and practices of interaction and exchange that take place within customer and supplier relationship and which need to be managed in order to develop successful co-creation opportunities (pp. 85-86)." Some encounters—those with more economic or emotional involvement—are more important for value co-creation processes, while other encounters may be trivial (Gremler, 2004). We therefore expect contests to be most effective when the task consumers are asked to complete is aligned with the brand positioning, including the core product attributes, their claimed consumer benefits and other higher-order associations. For example, Starbucks is focused on the perception that they offer the highest quality coffee and allow their customers to customize it. They also provide a warm atmosphere that becomes a home away from home for loyal customers. We predict that a contest inviting participants to discuss how they customize their cup of coffee or the role their favorite Starbucks store plays in their lives would outperform a contest in which participants had to do something unrelated to the brand idea, such as having to text or email Starbucks to be entered into a raffle to win a car. Writing about how Starbucks makes “their” special cup of coffee or how a visit to the local Starbucks provides a much-needed respite from an otherwise chaotic day forces consumers to think about Starbuck’s brand. Contests about cars would make no sense for Starbucks. Likewise, we predict that the Kit Kat contest, giving away NFL merchandise, would not be an effective brand builder, since the NFL has at best a weak association with the Kit Kat brand associated with taking a break. Sony would seem to be missing an opportunity by not having entrants write something about why they want the electronic devices. The Dove and Pepsi contests should have a strong effect on those creating the ads because these contests give the contestants an opportunity to actively think about the meanings and benefits they associate with the brand.
Empirical Test of Hypothesis
We analyze data from the Air Miles Reward Program (AMRP), which has been operating in Canada since 1992 and is one of the largest loyalty programs in the world, with over 10 million members representing over 67% of Canadian households. As a coalition loyalty program, members collect miles at over 100 sponsors categories covering all aspects of purchases ranging from groceries to gasoline to apparel and credit card purchases. Collected miles can be exchanged for rewards ranging from travel to merchandize to discount coupons.
In March 2009 AMRP launched a social media website for members to discuss the program and benefits. Member posts can be linked to their purchases, providing a unique opportunity to measure the effect of social media forums on purchase behaviors. Figure 1 shows the distribution of the number of posts per day during 2010. Posting is sporadic, with few posts on most days–the median number of posts per day is only 7 and the third quartile is 27. There are, however, large “spikes” in activity: the maximum number of posts on a single day was 6465. The spikes are driven by email promotions, e.g., on 2/3/10 AMRP sent an email announcing the one-year anniversary “Block Party,” in which members were offered up to 4 chances to win 25,000 reward miles by becoming a member, posting a tip, uploading a picture, or making an “I like this” thumbs up to the community site. This contest gave an opportunity for members to share their personal experience with AMRP or listen to other member’s experiences, which transforms customers from “passive audiences” to “active players” (Prahalad & Ramaswamy, 2000). The second contest, “Cruise,” culminated in a drawing to win a Caribbean cruise package for two people. During the span of six weeks, entrants had to answer six different questions each week, e.g., why they want to spend their week aboard or whom they want to take with them. The third contest, “Winter,” offered members a small number of miles for discussing what they planned to redeem their miles for in winter. This contest forced members to think about the core benefit of AMRP, and what rewards they want. The reason for belonging to AMRP becomes prominent in members’ minds, and the reward becomes more tangible by writing it down. There are some other much smaller spikes for less-successful contests that will not be analyzed in this paper.
AMRP provided the mile accumulation history for a stratified sample of 143,000 members, with all participants and a random sample of non-participants. Contest participants form the treatment group and non-participants are the control group. Having a control group enables us to control for threats to internal validity such as history. Pre-measures account for customer heterogeneity when comparing those who participate with those who do not. Although members self-select into participating in the contests, this before-after-with-control-quasi-experimental design is robust to most threats to internal validity (e.g., Churchill & Iacobucci, 2007). We also have more detailed information about one of the contests, which will be discussed in more detail below.
We analyze each contest separately. Black intervals on the horizontal axis of Figure 1 show study periods, defining the independent variable: for a contest, members who posted at least once in the study period are in the treatment group while others are in the control group. Let dummy variable xi equal 1 if member i posted something during the study period and 0 otherwise. For each contest the study period is labeled “1” above each of the black intervals on the horizontal axis. Pre-measures of behavior are computed from the 4-week period immediately preceding each study period, called the pre-period and labeled “0.” To study how long participation affects behavior in the future there are 3 additional periods labeled periods 2, 3 and 4. Study periods were selected to include most of the posting activities around a contest. The study period for the Block Party contest was 2 weeks long. The Winter contest was much shorter, with most posts occurring within a 1-week interval. The Cruise contest was longer, and its study period is 6 weeks. The pre-periods (0) were all four weeks long. The post periods were 2 weeks long for all but the Winter contest, where the post periods were 1 week long. Period lengths were round weeks because we suspect that accumulation behavior is at least somewhat periodic, with, for example, some households doing their grocery shopping every Saturday, etc. The dependent variables are number of miles accumulated during periods 1–4, yit for customer i during period t=0, 1, 2, 3, 4. We use a multiplicative model:

log(yit+1) = α + β1 xi + β2 log(yi0+1).

The log stabilizes the variance and symmetrizes the distributions. We test H0: β1=0 to decide if there is a difference between posters and non-posters. When we can reject H0 then we have evidence that posting during the study period affects behavior. Since this is a multiplicative model, exp(β1) describes how many times greater mile accumulation is for posters.
The results are provided in Table 1. In total, 12 regression models were estimated, one for each time period-contest combination. The number of participants is indicated under the contest name. The effect of participation in the Block Party contest on behavior during the study period (1) is β1=0.29435, which is statistically significantly different from 0. The magnitude of this effect is exp(0.29435)=1.34, implying that the accumulation behavior of those who participated in the Party contest was about 34% higher than those who did not. The effects in periods 2 and 3 are also significantly different from 0, but slightly smaller. Behavior is changed for a total of six weeks (three periods). This is a long-lasting effect. The results are similar for the Cruise and Winter contests. The effects of both contests are significant and substantial for all four periods.
We have more complete data from the party contest, with the email promotions executed before the contest began. The addition of the email data enabled us to estimate more accurate estimates of participation effect by controlling for the effects of email promotions. A total of 3,656,105 members were sent email invitations to participate in the contest. We also know which members opened the email, clicked on the link bringing them to the community website, and participated in the contest.
We have performed a simple analysis, as shown in Figures 2-4, by dividing our sample into 3 experimental blocks according to the number of miles they earned during the pre-period prior to the contest. The "Low" block earned less than 10 miles during the pre-period, the "Medium" block earned between 10 and 55 miles and the "High" block earned more than 55 miles. Each block constitutes approximately one third of our sample. These blocks control for differences in commitment to AMRP prior to the contest.
Consider Figure 2 for the Low customer block. The tree structure shows the number of customers and the mean number of miles accumulated during the contest period. For example, there were 7,846 people in the sample who did not receive the email and accumulated a mean of 6.2 miles during the contest period; 34,889 opened the email and accumulated an average of 6.1 miles, which is not significantly different from 6.2. Of those who received the email, 19,424 did not open the email while 15,465 opened it. Those who opened it accumulated 7.09 miles on average in the contest period, while those who did not open it accumulated 5.4 miles, which is significantly lower (P<.0001). Those who opened it are further split out into clickers versus non-clickers. Clickers are split between those who participate in the contest and those who don't.
In comparing the trees for the three blocks, we note four points. First, there is an accumulated effect as the customer becomes more engaged (i.e., from opening the email to clicking on the email to participating in the contest). Those who are open are significantly better than those who don't, those who click are better yet, and those who participate are even better. Participation seems to have a stronger effect than opening or clicking. Second, the absolute lift is greater for High customers (108.21-91.1=17.11) than for Low customers (17.9-7.8=10.1). Third, the relative lift of participation is greater for "Low" customers (10.1/7.8=129%) than for "High" customers (17.11/91.1=19%) (this interaction is significant). Fourth, High customers are more likely to engage in all three activities (open, click and participate) than other blocks. We have also estimated regression models that avoid the discrete blocks introduced by the above analysis, allowing us to control for other differences between the groups, and include interactions between pre-period characteristics and participation. The conclusions from these models are consistent with those from the trees presented above. Space does not permit us to present the models here.
This analysis provides a stronger test of the hypothesis that participation increases accumulation behaviors. In the earlier models participation was confounded with opening and clicking. Here we show that within those who clicked, participation still has an effect.
We have examined 3 social media contests and shown that those who participate consistently have significantly higher accumulation of miles during and after the contest. The longevity of the effect impressed us and we theorize that it is due to the participant co-creating the benefit of AMRP—earning a reward. By having participants write about the reward they want or why they want it, the benefit becomes more salient in their minds. Such contests also allow for the customization of the benefit, where participants can think about the reward that has unique meaning for them. In particular, the Cruise contest asked customers to share their personal stories about the meaning of the benefits they'll earn if they won the contest. These opportunities seem to be an example of "critical encounters" in that participants are encouraged to articulate the value of the brand promise in personalized contexts (Gremler, 2004). Such articulations should be a more potent form of advertising than merely exposing participants to an ad message telling them they can earn a reward.
This study has limitations and raises other questions. The quasi-experimental design used here, with pre-measures and a control group, is fairly robust to threats to internal validity, but it would be desirable to run true experiments where participation is manipulated, and to isolate the co-creation cause. Are contests where the task required for entry is aligned with the brand more effective than those that are unrelated? Is it important for participants to share their goals and experiences in a public forum, or is co-creating the benefit in private sufficient? Answering such questions will require experiments. We analyzed the effect of participating in contests, but could not examine the effect of reading entries made by others. Does reading a contest entry written by another customer change purchase behavior? Future research should address this issue to present a complete picture of how contests work for both active (i.e., creating messages) and less active (i.e., reading other people's messages) forms of participation.
There are important implications and applications of this research. First, out models enable ROI calculations for the contests. For example, rough ROI calculations can be produced from Figure 4 (High). The 1442 who participate accumulate 108.21-91.1 = 17.11 more miles, which means that AMRP issued 1442(17.11) = 25,000 incremental miles due to participation among the High customers. There are incremental miles from clicking and opening too. The total incremental miles multiplied by the amount AMRP charges its sponsors gives incremental revenue. We must do the same for subsequent time periods and the other two blocks. We would then deduct the cost of the promotion (sending 3.66 million emails) and the contest prize (25,000 miles).
Another implication of the co-creation explanation is that companies should design contests that force participants to focus on the core intended brand benefits. For example, the Kit Kat give-away contest should be more effective if entrants discussed how core brand associations were meaningful to them, e.g., “tell us how you take a break with Kit Kat,” or “tell us how you like to share your Kit Kat with a friend.” In contrast, filling out a form to win an NFL beverage pail would not seem to build the brand.
Allowing consumer to co-create brand meaning and benefits would seem to be a more effective way to increase purchase behavior than simply exposing consumers to ad messages, but it also creates new challenges and risks. The meanings that some consumers create may not be “on strategy,” in that they differ from what the organization intends for consumers to think about its brand. In some cases these alternative interpretations of the brand may be opportunities for the firm to expand the meaning of its brand. The firm may not have realized that this alternative meaning is the reason why a large number of consumers are loyal. It is also possible, however, to envisage situations where the alternative interpretations cause the brand to lose focus and confuse other customers who read them. Famous examples of such user-generated content discussed by Deighton (2008) are the Dove Cream Oil Body Wash entries by “Biker Dave” and “Bed Vlog.” Such ads are still available on YouTube and do not communicate the intended brand meaning. An important opportunity for future research is in devising communication strategies for responding to such user-generated content.

Figure 1: Distribution of the number of posts over time.

Figure 2: Tree diagram for "Low" customer block.

Figure 3: Tree diagram for "Medium" customer block.

Figure 4: Tree diagram for "High" customer block.

Table 1: Parameter estimates of β1, indicating the effect of participating in the contest. Below each estimate is the t statistic and the number of participants in each contest . One star indicates P<.05, two stars mean P<.01, and three stars mean P<.001.


Period 1

Period 2

Period 3

Period 4


n1 = 3,639/132,307


1.34 (12.82)


1.26 (10.18)


1.26 (9.63)


1.34 (12.71)


n1 = 19,473/136,237


1.23 (21.93)


1.16 (12.82)


1.13 (10.88)


1.11 (9.04)


n1 = 6,801/129,589


1.28 (13.28)


1.00 (-0.21)


1.05 (2.74)


1.10 (5.49)

Churchill, G.A. and Dawn Iacobucci (2008), Marketing Research Methodological Foundations, 9th edition, Thompson.

Deighton, J. (2008). “Dove: Evolution of a Brand,” Harvard Business School case number 9-508-047 and teaching note.

Gremler, D. D. (2004). The critical incident technique in service research. Journal of Service Research, 7(1), 65-89. doi: 10.1177/1094670504266138

Payne, A. F., Storbacka, K., & Frow, P. (2008). Managing the co-creation of value. Journal of the Academy of Marketing Science, 36(1), 83-96. doi: 10.1007/s11747-007-0070-0

Prahalad, C. K., & Ramaswamy, V. (2000). Co-opting customer competence. Harvard Business Review, 78(1), 79-87.

Prahalad, C. K., & Ramaswamy, V. (2004). Co-creation experiences: The next practice in value creation. Journal of Interactive Marketing, 18(3), 5-14. doi: 10.1002/dir.20015

Vargo, S., &Lusch, R. (2004). Evolving to a new dominant logic for marketing. Journal of Marketing, 68(1), 1-17.

Download 41.4 Kb.

Share with your friends:

The database is protected by copyright © 2024
send message

    Main page