Segmentations remain one of the most popular types of quantitative research studies, and when done right, they have the potential to transform an organization’s marketing strategy (including personas, messaging, and deliverables), especially from a product or brand portfolio perspective. As part of our extensive work on developing Kelton’s Segmentation Journey approach, the Kelton team has revisited one of the trickier aspects of a segmentation strategy—development of the typing tool.
What is a segmentation typing tool and what is its role?
As Orme and Johnson have succinctly defined, a typing tool is “an abbreviated set of questions for classifying new respondents into existing segments.” From an operational standpoint, a typing tool comprises a questionnaire and an algorithm designed to define customer segments. Some typing tools are built into excel spreadsheets. From there, qualitative inputs (from a segmentation survey) are coded and scored to define segments.
While this all sounds straightforward in theory, it represents a number of formidable challenges in reality. Many of these challenges have become more prominent recently given the growing demand for the integration of (survey-based) segmentation schemes into client databases, especially in a “Big Data” environment.
During the course of the many segmentation studies we work on, we are constantly exposed to these challenges and do unfortunately encounter unrealistic expectations regarding what a typing tool can and can’t do. In fact, clients are often surprised to discover that their typing tool needs (presumably a non-sexy formality and handled via some quick “back-end statistics”) have the potential to completely change the overall design of an in-depth segmentation solution.
To help align expectations and set our clients up for success, we therefore candidly discuss the following 3 Typing Tool Truisms at the very outset of a project:
Reality Check #1: Nuance is the enemy of typing tool efficiency.
Clients often crave market segmentations that shine with detailed attitudinal or psychographic nuances and communicate this desire with the archetypical “make them come to life” request. This is doable, but what tends to be forgotten is that the more nuances there are in your data set, the more difficult it gets to create an accurate typing tool. It’s a tradeoff that is logical, yet often conveniently ignored until low typing tool accuracy numbers are reported.
- The Takeaway: Tackle the White Elephant right away. No one will prefer a simple solution over a nicely nuanced solution unless you reinforce the idea that the accuracy rates will drop significantly with growing nuance.
Reality Check #2: In segmentation research, the definition of “good” accuracy varies…
There are no hard and fast rules for which typing tool accuracy levels are awesome, mediocre, or unacceptable (try googling it), and one of the reasons for that is that it very much depends on a segmentation’s complexity. Sure, you can achieve an accuracy of 90% or higher if your segmentation research is based solely on a couple of demographics and maybe a core purchase spend number. But once you add, say, 10 attitudes, a couple of bipolar trade-off statements, some tech ownership variables and a few shopping preferences to the input mix, be prepared for accuracy levels to drop into the 70s, 60s, or even lower—and that’s “good” given the complexity.
- The Takeaway: Have a candid conversation about both the desired segmentation complexity and the desired typing tool accuracy—and make sure they match up.
Reality Check #3: You can’t predict apples with oranges. Really.
This is where typing into databases comes into play. Now, maybe your customer database has 10,000+ variables for each record, including some attitudinal metrics—consider yourself lucky and jump to the final paragraph; however, the reality is that most databases have a modest amount of variables in data points, many of which are strictly behavioral (consumer spend, website visitation frequency, etc.) along with some basic demographics. So how well do you think these variables are at predicting consumers’ needs, feelings, attitudes, and preferences? Usually not good at all, and if done wrong, your segment membership will be, too.
- The Takeaway: Don’t beat around the bush, but address this challenge head on. If database typing is even a remote possibility, start the project by learning everything you can about the database and alert the client to what that means for potential typing tool accuracy (of course, also depending on the desired segmentation complexity).
Are you exhausted yet? Yes, it’s a lot to consider and there are no easy solutions for any of it. Plus, it means you have to discuss limitations and challenges instead of the beauty of the distinct segments you will create and nicely visualize. Awesome. But is has to be done since you’ll be doing both yourself and your client a favor… Which brings me to my last and most important piece of advice.
With segmentation research, always start with the end in mind!
You must find out what the client’s expectations are for post-segmentation typing tool applications and then design the segmentation approach and methodology accordingly, from the ground up. Should it be a deep attitudinal segmentation, an occasion-based segmentation, or a demographics-dominated segmentation? In extreme cases you might even recommend to not use survey research at all and create the segments directly off database variables. And all that just because of this little thing called the typing tool.
Finding the right partner can make all the difference in the success—and effectiveness—of your segmentation research studies. You’ll want to work with an agency with adept machine learning capabilities, best-in-class segmentation design and segmentation analysis, and the know-how to conduct proper validation of findings.
When you’re ready to partner, contact Kelton (now Material) by filling out this form.