Many segmentation studies share this fatal flaw-Don’t launch your next segmentation until you read this.

By Joe Beier

“Behold the power of segmentation done right”

Segmentation models go a long way to solving a universal business challenge- how to most effectively deploy always limited support resources for maximum marketplace impact.  And identifying what really matters to the groups of consumers that really matter most to us?  Vast marketplaces are broken down into manageable and targetable segments.

“Agree to disagree”

The fatal flaw that limits the effectiveness of most segmentation studies is a reliance on “agreement scale” questions.  Respondents are typically asked to rate how much they agree /disagree that a given statement about attitudes or behaviors describes them.  While scaled questions are widely used, they are also widely recognized as being plagued with some key limitations that become particularly problematic in the context of a segmentation including:

• Producing a stated (more prone to human memory limitations) vs a behavioral (not memory reliant) data point
• Lower discrimination across respondents as possible responses now limited by “ends” of scale.
• Distortion introduced by respondent tendency to “fixate” all responses near one end of scale
• Poor consistency across geographies/cultures (i.e. systemic bias to rate everything high or low)

In sum, it is our belief that reliance on these “agreement scale” questions is largely responsible for the high level of client dissatisfaction with segmentation studies vs other common research solutions.  Even the most brilliant post-field analytics cannot salvage a study whose front-end inputs are so deeply flawed.  The “GIGO” rule proves true once again!

A better (and Behavioral) data collection method

Building better front-end data inputs starts with deep-dive cognitive interviews that illuminate the true, and sometimes sub-conscious motivators driving consumer attitudes and behaviors.  Our team then translates these qualitative drivers into discrete statements that can be tested quantitatively.  (i.e. “It is better to manage my health holistically through lifestyle than to rely on Rx medications”) Statements are then exposed in a discrete choice exercise that forces respondents to make choices between them in selecting those that “best apply to them”.  These behaviorally-based data sets then feed into our sophisticated statistical modeling (Bayesian latent class) to produce superior segmentations.

“Proof in the pudding”

Clients love segmentations built with this front end qual and discrete choice respondent experience because:

• They are defined by the deep and true drivers of consumer beliefs and behaviors vs more superficial variables like “level of category engagement”.
• Segments are derived more holistic and not bound by a “scale question” continuum
• Segments are intuitive and meaningfully distinct from one another
• Targeting and activating against segments is highly efficient as there is extraordinary homogeneity inside each segment.

Not all segmentations are created equal. For results like these, please talk to us before you embark on your next segmentation journey.