The New Lynchpin of Targeted Marketing: Correlation

Spock, the character played by Zachary Quinto in Star Trek Into Darkness, emphatically notes “the needs of the many outweigh the needs of the one.” While this quote may be apt for a Starfleet captain, or even a first officer, it does not necessarily resonate for marketers. A key aspect of today’s digital world is the ability to target an ad or an offer to the individual customer—the segment of one—with the explicit idea of catering to the needs of the one.

While direct marketers have done targeted marketing for a long time, the availability of online behavior of individual consumers along with their purchase outcomes has allowed marketers to zero in on the individual customer and to “scale up” this targeted marketing effort to a large number of (potential) customers. Amazon, for example, can provide us with recommendations for books and movies on the basis of our specific purchase patterns overlaid on those of others who make the same purchases. Health-care websites recommend doctors on the basis of some characteristics we provide about ourselves. And we are able to design our own customized iMac on the Apple website.

The key to traditional market segmentation has been in the identification of differences across customers (or customer groups) along some behavioral dimension or dimensions relevant to the marketer. For example, if I market chewing gum, I would be interested in those customers who chew a lot of gum. Segmenting my customers by their usage rate allows me to make decisions regarding whether I would like to focus my marketing activities on those with a high level of usage or those with a low level of usage.

Next, I need to describe these segments: Are the heavy-user consumers young or old? Do they have high incomes or low incomes? Without describing the segments, I would not be able to appropriately target a segment chosen on the basis of its “attractiveness” on the usage dimension. Without describing the segments, how would I know who the heavy users are, which media they follow, and what channels they shop in? Once I know that my targeted heavy-user segment consists of older consumers, I can pick the appropriate media they pay attention to and the channels they shop in.

The last piece is that of positioning, i.e., convincing customers in my target segment to purchase my product—a topic that we have covered elsewhere.

Two recent articles, one in the Wall Street Journal and the other in the Economist, present an interesting twist on this traditional view of segmentation and extend it to the digital realm. The first article looks at research that correlates patterns of behavior on social networking sites with the demographic and psychographic profiles of the users of those sites. For example, researchers found that Facebook “likes” for words such as swimming, chocolate-chip-cookie-dough ice-cream, etc. correlated with individuals who did not use drugs. Importantly, these analyses are able to look at “likes” and uncover whether users are Democrats or Republicans; black or white; homosexual or heterosexual, etc. with a high degree of accuracy. The second article discusses research at IBM’s Almaden research center, where researchers looked at individuals’ Twitter streams and identified words that then correlated with those individuals’ psychological profiles. For example, the word summer when mentioned in tweets correlated with the characteristic of trust.

Such analyses are interesting because they help us look at some element of a user’s behavior (e.g., Facebook “likes”) and then infer that person’s demographic and psychographic profile. But why is this of interest to marketers? The answer is simple: using information on individuals’ online behaviors, I can link one type of behavior (say, “likes”) to other types of behavior (taking a vacation in the Bahamas). What this enables me to do is to relate taking vacations in spots like the Bahamas to the demographic and psychographic characteristics of those individuals. Once I do that, I am back to my targeting stage—armed with information on descriptor variables (demographics, psychographics, etc.) that relate to different types of behaviors, I can pick media and channels to direct my marketing effort appropriately.

A note of caution while using such methods is that they are based entirely on correlational patterns. Often, such methods are useful for descriptive purposes but do not establish causal linkages between, say, advertising directed at the chosen group and the outcomes in terms of actual booking of vacations. So one should be careful to recognize that more would have to be done before using these analyses for purposes of decision-making. As Karl Urban, Dr. Leonard “Bones” McCoy in the aforementioned Star Trek movie, wryly notes, “When I dreamed about being stuck on a deserted planet with a gorgeous woman, there was no torpedo.”

Pradeep K. Chintagunta is the Joseph T. and Bernice S. Lewis Distinguished Professor of Marketing at Chicago Booth.