Bad Thing! Or Why Segmentation by Consumer Attitudes May Be Dangerous

For years, B-to-B and B-to-C marketers have relied on attitudinal segmentation research to help them group their current customer base, and potential customers as well, for communication, promotion, marketing and experience initiatives. The thesis has been that, by asking a small, but meaningful, set of attitudinal questions, they would be able to develop an index, algorithm or framework equation that ranked these consumers by propensity to buy, both near-term and long-term.

For years, B-to-B and B-to-C marketers have relied on attitudinal segmentation research to help them group their current customer base, and potential customers as well, for communication, promotion, marketing and experience initiatives. The thesis has been that, by asking a small, but meaningful, set of attitudinal questions, they would be able to develop an index, algorithm or framework equation that ranked these consumers by propensity to buy, both near-term and long-term.

These frameworks—they’re arithmetic, so we can’t call them “models”—typically include questions regarding the importance of elements like value for money, acting with the consumer’s interests in mind, credit and payment terms, having knowledgeable employees, offering products which will meet the consumer’s needs, and the like. From these questions, basic segment categorization can be determined; and, once these three, four or five segments are established, we’ve often seen marketers go on to build assumptive plans and conduct further, more detailed, research around them.

The goal of these approaches is to produce attitudinal segments, which the questions can predict with high accuracy, often in the 80 percent or 90 percent range. This creates what economists would call a “post hoc ergo propter hoc” situation, Latin for “after this; therefore, because of this.” It is a logical fallacy, essentially saying that A occurred (the responses to the attitudinal questions); and then B occurred (the cuts, or segments, of consumers). Thus, A caused B. Once the B, or segment creation, stage has been established, further fallacies, such as creating reliable marketing, operational and experiential strategies around these supposed propensities, can be built. It’s a classic situation, where correlation is thought to be the same as causation. As your economics or stat professors may have told you, correlation and causation are far from being identical concepts.

As a consultant and analyst, I’ve seen this result of this application of research and analytics play out on a firsthand basis on multiple occasions. Here’s a recent one. A client in the retail office products market had been using an attitudinally derived element importance question framework for small business market segmentation purposes. The segment assumptions went unquestioned until followup qualitative research was conducted to better shape and target their planned marketing and operational initiatives. Importance of certain products and reliable service were identified in the research as key areas of focus and opportunity for the office products retailer; but, in the qualitative research, power of both focus areas appeared, anecdotally, to be consistent across all segments. And, even though implied supplier roles were suggested to build purchases, this was much more “leap of faith”-based on the established quantitative research segment personas than actual qualitative research findings.

There are related issues with what we can describe as quasi-behavioral measures, such as single question metrics (likelihood to recommend to a friend or colleague or the amount of service effort required on the part of a consumer); or traditional customer loyalty indices (where future purchase intent is included, but also attitudinal questions such as overall satisfaction). It’s not that they don’t offer some segmentation guidance. They do—on a macro or global level; but they tend to be less effective on a granular level, especially where elements of customer touchpoint experience are involved.

And, they tend to have limitations as predictors of segment behavior, a key business outcome for marketers and operations management. When compared to research and analysis techniques, such as customer advocacy and customer brand-bonding, which are contemporary, real-world frameworks built on actual customer experience—high satisfaction scores, high index scores and high net recommendation scores produced likely future purchase results (in studies across multiple industries) which were often 50 percent to 75 percent lower than advocacy or brand bonding frameworks. I’d be happy to provide proof for anyone interested in reviewing the findings.

So, that’s the scenario. The challenge, and potential danger, for marketers and those responsible for optimizing customer experience is that these attitudinal and quasi-behavioral questions are just that—attitudes and quasi-behaviors. Attitudes are fairly superficial feelings, and tend to be both tactical and reactive. And, because they are so transitory, their predictive value is often unstable and unreliable. Quasi-behaviors are also open to many similar challenges. More importantly, attitudes and quasi-behaviors are not behaviors, such as high probability downstream purchase intent based on actual previous purchase, evidence of positive and negative word-of-mouth about a brand based on prior personal experience, and brand favorability level based on experience. These are especially valuable in understanding competitive set, and they have real, and very stable, predictive and analytical value for marketers.

As Jaggers, the lawyer, said to Pip in Charles Dickens’, “Great Expectations,” take nothing on its looks; take everything on evidence. There’s no better rule.” For marketers, that’s excellent shorthand for taking everything on behavior, and perceptions based on documented personal experience, rather than attitudes and quasi-behaviors.