«1 How Interest Shapes Word-of-Mouth Over Different Channels JONAH BERGER RAGHURAM IYENGAR* * Jonah Berger is the James G. Campbell Jr. ...»
How Interest Shapes Word-of-Mouth Over Different Channels
* Jonah Berger is the James G. Campbell Jr. Assistant Professor of Marketing
(email@example.com) and Raghuram Iyengar is an Assistant Professor of Marketing
(firstname.lastname@example.org) at the Wharton School, University of Pennsylvania, 700 Jon M.
Huntsman Hall, 3730 Walnut Street, Philadelphia, PA 19104. Author order is alphabetical. Pete Fader, Dina Mayzlin, Renana Peres, Andrew Stephen, and Christophe Van den Bulte provided helpful comments on earlier versions of the manuscript. The authors greatly appreciate the data provided by Ed Keller and the Keller Fay Group. The research was partially funded by the Wharton Dean’s Research Fund.
There has been a great deal of recent interest in word-of-mouth, but while it has been shown to boost diffusion and sales, less is known about its causes, or what leads people to talk about certain products or brands rather than others. Further, consumers share word-of-mouth through different conversation channels (e.g., online vs. face-to-face), but do these channels shape what types of products and brands get discussed, and if so, how? This research addresses this question, providing insight into how the channel people share word-of-mouth through impacts what gets discussed.
Consumers share word-of-mouth face-to-face, online, and through various other channels. But do these channels affect what people talk about, and if so, how? Analysis of over 21,000 conversations, as well as a laboratory experiment, demonstrate that conversation channel continuity norms shape what gets discussed. In discontinuous conversation channels (e.g., online posts or text), pauses between conversational turns are expected, so people have time to select and craft what they say. Consequently, more interesting products should be talked about more than boring ones. In channels where conversations are expected to occur more continuously (e.g., face-to-face or on the phone), however, there is less time to selectively pick what one talks about. Consequently, how interesting products are to talk about should have less of an impact on whether they get discussed. These findings shed light on what drives word-of-mouth and how companies can design effective word-of-mouth campaigns.
Word-of-mouth is frequent and important. Consumers talk about restaurants they like, post reviews of movies they hate, and share information about the best child safety seats. Social transmission also has a significant impact on what people buy and how they behave (Godes et al.
2005; Godes and Mayzlin 2004; 2009; Iyengar, Van den Bulte, and Valente 2011; Leskovec, Adamic, and Huberman 2007). Consequently, organizations have come to realize that generating word-of-mouth is an important part of marketing strategy.
But while its consequences are clearly valuable, much less is known about word-ofmouth’s causes, or what leads people to talk about certain products or brands rather than others.
Further, the little work in this area has mostly ignored whether the conversation channel shapes what people talk about. Word-of-mouth can be shared in different ways. People have face-toface conversations, post on blogs, send texts, or write online reviews. Do these different channels shape what types of products and brands people talk about, and if so, how?
Consider the brands that get talked about most over different channels. Virtue.com does an annual ranking of which brands are talked about most online and The KellerFay Group provides a similar index for offline word-of-mouth. Comparing the top 10 brands on each list, however, shows little overlap. Verizon is 2nd in offline word-of-mouth, for example, but ranks 85th in online conversations.
While these differences might reflect different methodologies used by the two firms, or differences in people who talk more over one channel versus another, might they also say something deeper about the psychology behind word-of-mouth?
This paper distinguishes between different types of conversation channels (i.e., continuous and discontinuous) and uses this notion to shed light on one way in which the channel itself shapes word-of-mouth. In particular, we examine how conversation channel continuity moderates the relationship between how interesting a product or brand is to talk about and how much word-of-mouth it receives. We do this in two ways. First, we analyze two unique datasets of thousands of everyday conversations across different channels to provide evidence for our theoretical perspective in the field. Second, building on these results, we conduct a controlled laboratory experiment where we manipulate conversation continuity to directly examine its effects. The results underscore the causal impact of channel continuity on word-of-mouth.
Taken together, the results deepen understanding about what drives word-of-mouth and provide insight into how to design more effective word-of-mouth marketing campaigns.
Most research on word-of-mouth has focused on how it affects diffusion and sales.
Consumers are more likely to buy DVDs their friends recommend (Leskovec et al. 2007) and doctors are more likely prescribe new prescription drugs that other doctors they know prescribed previously (Iyengar et al. 2011). Similarly, word-of-mouth and online reviews have been shown to boost new customer acquisitions (Schmitt, Skiera, and Van den Bulte 2011; Trusov, Bucklin, and Pauwels 2009) and increase sales in various product categories (Chevalier and Mayzlin 2006; Godes and Mayzlin 2009). Certain types of word-of-mouth (i.e., explaining language) have even been shown to impact evaluations of consumption experiences (Moore 2012).
But while research has focused on the consequences of word-of-mouth, there has been much less attention paid to its causes (Cheema and Kaikati 2010; Goldenberg, Libai and Muller 2001), or how different communication channels (e.g., face-to-face vs. online vs. phone) shape what people talk about or share. Most papers rely on data from only one channel, such as online reviews (Chevalier and Mayzlin 2006), newsgroups (Godes and Mayzlin 2004), email forwards (Berger and Milkman 2012), email referrals (Leskovec et al. 2007; Trusov et al. 2009) or face-toface communication (Berger and Schwartz 2011; Cheema and Kaikati 2010; Godes and Mayzlin 2009). But when only one channel is examined, it is difficult to say much about how the channel itself impacts behavior. Indeed, researchers have noted that there may be fundamental differences between online and offline social interactions (Godes et al. 2005), yet little research has addressed this point.
This issue is particularly important given that groups that want to increase word-of-mouth must decide which channel(s) they want to target. They choose whether to try and create a viral video, encourage online referrals, stage a flash mob, or generate some other event, promotion, or campaign to increase mentions of the brand. But these approaches are designed to encourage word-of-mouth through different channels. Consequently, to understand how to make them effective, we need to understand the nature of the channels themselves and whether they have different impacts on what gets shared.
We distinguish between continuous and discontinuous conversations and use this notion to shed light on how conversation channels shape conversation.
People communicate information when they talk, but as with many types of consumption behaviors (Levy 1959), they also communicate things about themselves (Tannen 2005; Wojnicki and Godes 2010). If someone quotes Shakespeare and Thoreau, people may assume they are well-read. If someone always talks about restaurants that just opened, people may infer that they are a foodie. Along these lines, Wojnicki and Godes (2010) show that consumer propensities to talk about satisfying and dissatisfying experiences depend in part on their desire to communicate domain expertise.
But people not only communicate through what they talk about, they also communicate through how they talk. Tannen (2005) notes that stylistic elements of conversation, such as rate of speech, speed of turn taking, and avoidance of pauses between conversational turns, all communicate things about the speaker. Failures to live up to expectations on these different dimensions can lead others to make negative attributions about a person (Loewenstein, Morris, Chakravarti, Thompson, and Kopelman 2005). Transitions from one party speaking to the other, for example, usually occur with no long gap or silence in between. Consequently, people who take long to respond may be seen negatively (Clark 1996; Sacks, Schegloff, and Jefferson 1974;
Expectations about conversation style, however, vary based on the conversation channel (e.g., face-to-face vs. email). Different types of conversations come with different norms (Grice 1975; Levinson 1983). Think about the last time you had lunch with a friend or shared a cab with an acquaintance. Most face-to-face settings as well as phone conversations involve continuous discussion (Sacks et al. 1974). There is an expectation that ongoing conversation will occur, and it is awkward to sit in silence. Both parties try to keep the conversation flowing, filling the conversational space, and discussion is relatively continuous with few breaks in between. Long pauses are somewhat uncomfortable and may lead people to infer that someone is not a good conversationalist.
Contrast that, however, with the types of conversations that often occur in online discussion forums, like blogs or Facebook, that are mostly discontinuous in nature. One person writes a post or comment, but there is no expectation that someone else will respond right away.
In fact, even if a person does decide to respond, it may occur hours or even days later. This is not only true of broadcast conversations (i.e., one-to-many like a blog post) but even in narrowcast or dyadic online conversations where only two people are involved. When someone posts on someone else’s Facebook wall, or sends them an email or text, they do not usually expect an immediate response, and even an “immediate” response is seen as one that occurs minutes later, rather than right away. Further, because the expectation is that conversation is asynchronous, people have time to deliberate and think through what they say. Text and email are similar. Overall, discontinuous conversations generally involve no expectation of immediate response, and pauses in the conversation are not seen to signal anything about the conversation partner.
Thus in an attribution sense, how strongly conversation partners will make inferences about the person they are talking to depends on the salience of a situational attribution. If Jason takes a while to respond, does it say something about him or just the medium over which we are conversing? If the conversation is over a channel that is expected to be discontinuous (e.g., email), people will be more likely to attribute a delayed response to the channel (e.g., he must not have received the note or had a chance to respond) and infer little about the person. If the conversation is over a channel that is expected to be continuous (e.g., face-to-face), however, people will be more likely to attribute a delayed response to something about the person.
We suggest that these differences in conversation continuity will impact the types of things that get discussed.1 In particular, we suggest that whether or not a product, topic, or brand is interesting to talk about will have a greater impact on whether it gets mentioned in certain channels (e.g., online) rather than others (e.g., offline), and that this is driven by differences in the conversation continuity of the channels.
frequently discussed potential drivers of word-of-mouth. Practitioners often argue that products need to be interesting (i.e., novel or surprising in some way) to be talked about (Dye 2000;
Hughes 2005; Knox 2010; Rosen 2008). In his popular book on word-of-mouth marketing, for example, Sernovitz (2006) argues that the most important way to generate word-of-mouth is to “be interesting” and that “nobody talks about boring companies, boring products, or boring ads,” (p. 6). Thus we test whether this common wisdom holds, and whether it holds equally, in different word-of-mouth channels (e.g., online and offline).
Second, prior work has found conflicting relationships between interest and WOM.
While theory suggests that more interesting products should be talked about more than less
Researchers have also described this difference in terms of synchronous vs. asynchronous communication (Poole, Shannon, and DeSanctis 1992), but we prefer to talk about conversation continuity (see Sasks et al. 1974) because it more concretely reflects the pauses that do, or do not happen between conversational turns.
interesting ones (Dichter 1966), and some empirical work supports this notion (Berger and Milkman 2012), other work shows that more interesting products do not get more word-of-mouth (Berger and Schwartz 2011).
We suggest that this seeming discrepancy in prior findings is at least partially due to differences in the expected conversation continuity of the different word-of-mouth channels examined. When conversations are expected to be discontinuous, people have time to select and craft what they say. They have more opportunity to think of a clever or interesting response and can wait to respond until they have something worthwhile to talk about. One author’s friend, for example, notes that he is much more suave over text than in person because he can take the time to craft the perfect response. Consequently, in discontinuous channels people should be more likely to post or share something if they think it will be interesting. Indeed, prior work shows that more interesting New York Times articles are shared more frequently online, and are more likely to make the Times most emailed list (Berger and Milkman 2012).