|
Article Excerpt
Managers are very interested in word-of-mouth communication because they believe that a product's success is related to the word of mouth that it generates. However, there are at least three significant challenges associated with measuring word of mouth. First, how does one gather the data? Because the information is exchanged in private conversations, direct observation traditionally has been difficult. Second, what aspect of these conversations should one measure? The third challenge comes from the fact that word of mouth is not exogenous. While the mapping from word of mouth to future sales is of great interest to the firm, we must also recognize that word of mouth is an outcome of past sales. Our primary objective is to address these challenges. As a context for our study, we have chosen new television (TV) shows during the 1999-2000 seasons. Our source of word-of-mouth conversations is Usenet, a collection of thousands of newsgroups with diverse topics. We find that online conversations may offer an easy and cost-effective opportunity to measure word of mouth. We show that a measure of the dispersion of conversations across communities has explanatory power in a dynamic model of TV ratings.
Key words: word of mouth; diffusion of innovations; measurement; networks and marketing; new product research; Internet marketing
History: This paper was received June 3, 2002, and was with the authors 13 months for 3 revisions; processed by Gary Lilien.
1. Introduction
Among the many and varied channels through which a person may receive information, it is hard to imagine any that carry the credibility and, thus, the importance of interpersonal communication, or word of mouth (WOM). There is little debate as to whether WOM matters to the firm. There is good reason to believe that it has more potential impact than any other communication channel. Katz and Lazarsfeld (1955) showed that WOM was the most important source of information for certain household items. More recently, a study by Jupiter Communications (1999) found that 57% of people visiting a new Web site did so based on a personal recommendation; this is higher than any other source of influence. As these studies suggest, managers are interested in WOM because it is often an important driver of consumer behavior such as the adoption of a new technology, the decision to watch a TV show, or the choice of which laptop to purchase. It might affect awareness in some cases, or preferences in others. Alternatively, WOM may simply serve as a leading indicator of a product's success. Whatever the specific mechanism, there is empirical evidence, as well as an intuitive justification, for the hypothesized link between WOM and consumer behavior.
One implication of this relationship is that the firm should measure WOM. As a leading indicator, WOM measurement would be important for market research. As a driver of behavior, WOM measurement would be a prerequisite to effective "buzz management." To paraphrase Edward Deming, "You can't manage what you can't measure." However, there are at least three challenges associated with measuring WOM. First, how does one gather the data? Because the information is exchanged in private conversations, direct observation has traditionally been difficult. As a result, most marketers and researchers either have relied on consumer recall or have inferred the process of information exchange from aggregate data. An important implication of the rise of online communities is the enablings of observation of consumer-to-consumer conversations. Here we investigate the potential use of these conversations in measuring WOM.
Second, even if we could observe the conversations, what aspect of them should we measure? How does one measure a set of statements between people? Which of the infinite transformations of a conversation are meaningful and managerially useful? The most common approach is to use simple counts. This approach is similar to news-clipping services that monitor how many times a firm's products are mentioned. We investigate the informativeness of this naive measure. We also investigate another dimension of WOM: dispersion. We define this construct as the extent to which product-related conversations are taking place across a broad range of communities. We expect that less dispersed WOM--discussions focused within a narrow and homogenous population--is likely to have less of an impact than broadly dispersed WOM.
The third challenge comes from the fact that WOM is not exogenous. While the mapping from WOM to future behavior is of interest to the firm, we must recognize that WOM is also an outcome of past behavior. This has important implications for the measurement of WOM. For example, high WOM today does not necessarily mean higher sales tomorrow. It may just mean that the firm had high sales yesterday. Thus, to understand the nature of the link, we need to understand the dynamic relationship between consumer behavior and WOM. Further, we must allow for the fact that the role and the impact of WOM may change over a product's life.
Our primary objective in this paper is to address these challenges. In so doing, we will evaluate the informativeness of two measures--volume and dispersion--to the manager. Specifically, we envision a manager attempting to learn from aggregate data the underlying process governing her customers' behavior. If she had the opportunity to measure WOM, we offer unique insight into which aspects of it she should measure. Given this focus, we are seeking measures that are practical to implement at reasonable cost. We make no claim concerning the optimality of the investigated measures. Another objective of the paper is to investigate the usefulness of online conversations in the study of WOM. The context we study is characterized by a decision made offline, yet we measure WOM online. Thus, to the extent that we find that certain measures are informative, we argue that this supports the idea that at least some aspects of online WOM are proxies for overall WOM. Given the operational advantages of measuring WOM online, we hope to spur a significant increase in focus on the Web as a laboratory for WOM research.
As a context for our inquiry, we study the relationship between TV viewership behavior and WOM. Specifically, we analyze the ratings for new TV shows during the 1999-2000 seasons. WOM appears to be especially important for entertainment goods: A recent Forrester report concludes that approximately 50% of young Internet surfers rely on WOM recommendations to purchase CDs, movies, videos or DVDs, and games (Forrester Research 2000). Note that the decision to view a TV show is made repeatedly. This is interesting because the consumer's purchase experience in period t will affect her decision to talk about it as well as her consumption decision in period t + 1. Our source of WOM information is Usenet, a collection of thousands of newsgroups with very diverse topics.
The paper proceeds as follows. After reviewing the relevant literature in [section]2, we discuss our research objectives in [section]3. In [section]4, we describe the two sources of data used in the study: Nielsen ratings and Usenet. In [section]5, we present the main empirical results. We find that higher WOM dispersion is related to higher future ratings. We also find that the impact of dispersion declines over time. This argues for measuring WOM early in a product's life. Surprisingly, we find that volume is not consistently associated with higher future ratings. We discuss this result in [section]6. One potential explanation for the null result could be the fact that positive and negative volume have offsetting associations with future ratings. Because the valence of the post is unobserved in our main analysis, these effects may cancel each other out. To test this, we collect valence information for a sample of the data. Nonetheless, this more costly analysis does not yield the expected association between the volume of WOM and future ratings. Another explanation might be that there is less additional information from a volume measure--as compared with dispersion--conditional on past ratings. Our three equation estimations (dispersion, volume, and ratings) provides some support for this. We conclude in [section]7 with a discussion of the findings, their implications, their limitations, and suggestions for future work.
2. Literature Review
Our work draws on three streams in the WOM literature: (1) WOM as a driver of buyer behavior, (2) the importance of social structure in the flow of WOM, and (3) WOM as an outcome of consumer behavior in the past. In addition, we discuss the traditional approaches that have been used to measure WOM.
2.1. WOM as a Driver
There exists ample theoretical support for the idea that WOM impacts consumers' actions. Banerjee (1992, 1993) presents two models that suggest that people are influenced by others' opinions. In fact, rational agents may ignore their own private information in favor of information inferred from others' actions. This may lead to "herding" in which all agents select the same action, which at times may be suboptimal. A similar context is analyzed by Bikhchandani et al. (1991). An important implication of the latter group's work is that the introduction of new information can cause discontinuous shifts in the actions of the agents. This may explain fads and bubbles. Mayzlin (2004) focuses specifically on WOM online and the potential that it presents for the firm to pose as a consumer and create firm-to-consumer communications that look like consumer-to-consumer communications. She finds that, even when this is possible, rational consumers still pay attention to anonymous online posts. As a result, posing as a customer online may be a profitable equilibrium strategy for the firm.
There have also been numerous experimental and empirical attempts to provide support for this role of WOM, with mixed success. Reingen et al. (1984) conduct a survey of the members of a sorority in which they measure brand preference congruity as a function of whether they lived in the sorority house. Those who lived together had more congruent brand preferences than those who did not. Presumably, living together provides for more opportunities for interaction and communication. Of course, because of the nature of the study, the authors cannot definitively rule out an alternative explanation that women with similar tastes choose to live together. A similar study, in a different context, was performed by Foster and Rosenzweig (1995). They look at the adoption of high-yield varieties (HYV) of seeds among Indian farmers. They find that the profitability of farmers employing the HYVs was higher as the adoption rate of the village increased. They interpret this as a learning spillover. Again, the presumption here is that there is significant WOM at the village level which facilitates the flow of information regarding the new technology....
|
