Learn how qualitative research can be more targeted and effective. Using segmentation analysis to create focus groups requires you to both understand the technical definitions of the segments and to develop statisical-based screening questions that select "pure types" for each segment.
Woe is me...
We recently finished a massive study to segment our market and went to do some focus groups using the segmentation analysis. The trouble was the respondents in the focus group did not seem to fit our segments. What happened? Was the segmentation wrong? We had an Excel spreadsheet that the recruiters were supposed to use to put the respondents in segments. Did the focus group recruiters not follow the screener?
The reasons why things like this happen can be murky, but two factors often result in outcomes like this:
- Vivid descriptors, and
- Fuzzy segments
A market researcher with any sense of marketing is unlikely to send a client the results of a segmentation study with segments named Segment 1, Segment 2, and Segment 3. Segments need vivid names; not boring names like Segment 1, Segment 2, and Segment 3, but names with some sizzle like Evaluators, Loyalists, and Defectors.
Of course, the natural outcome is that the labels take on a life of their own, often with much more meaning than the particular bundle of attitudes, behaviors, and demographic characteristics that actually define the segment.
To call a segment Evaluators is to evoke associations and attributes that may have had nothing to do with the segment definition. For example (and this is similar to an actual Yarnell research case), to call a segment Evaluators suggests an active, engaged person who has actually thought about multiple products in a category while in actuality all it might mean (because of how the segment was defined) was that they had at one time used a competing product. So, when the clients sat in the focus group observation room, respondents did not act the way the clients expected Evaluators to act.
What do you do about this? Make sure to discover the technicial definitions of every segment. More often than not, the vivid descriptor leads people to extend the meaning of the segment far beyond its technical definition.
A rose by any other name...
Be very careful when developing segment names, or when presented with someone else's segment names. The person who names the segment has an extraordinary influence on what the segments mean to your company, which significantly affects key marketing processes, such as product development and messaging. Make sure the names are clear and accurate and impress on the people who will use the segmentation the technical, full meaning of each segment.
Check Goodness of Fit...
A segmentation where many people have a moderate likelihood of being in multiple segments may not be particularly useful. So, before accepting any segmentation study, ask to see goodness-of-fit statistics. This will help you understand how many people had a high probability of being in only one segment (pure types) and how many had more-or-less equal probabilities of being in more than one segment. In other words, a goodness-of-fit statistic can help you figure out whether you have a satisfactory segmentation.
A fundamental rule of segmentation is that everybody has to be in a segment and nobody can be in more than one segment. Segments are typically derived from a collection of attitudes, behaviors, and demographic characteristics. Often segmentation uses cluster analysis to identify combinations of these attitudes, behaviors, and demographic characteristics that divide the market into clumps of people with similar characteristics and typically relate to some market behavior.
However, determining where everyone goes is not that simple. People are different and short of putting everyone in their own segment, any segmentation will be an approximation of where each person in a segment is more like (but not identical to) the other people in that segment than people not in that segment. The catch is the phrase in parenthesis.
Another way of thinking about this is that a segmentation analysis assigns each person a probability of being in each segment. People are assigned to the segment they have the highest probability of being part of. This does not mean, however, that everyone has a high probability of being in only one segment. They may have a moderate likelihood of being in multiple segments.
So, when screening for focus groups using a segmentation analysis like this, you often end up with a mix of people who fit the segment with a high precision (they had a high probability of being in only that segment) and people who do not (even though it was the best fit for them). For example, a respondent with 0.9 probability of being a Loyalist (and 0.1 of being an Evaluator and 0.2 of being a Defector) will fit into a Loyalist group with a high degree of precision, profiling very closely as a "pure" Loyalist. A respondent with 0.4 probability of being a Loyalist (and 0.2 of being an Evaluator; and 0.1 of being a Defector) will also end up in the same focus group despite the fact that they do not really profile as a pure Loyalist but fit in that group because that is the best fit for them.
What do you do about this when using the segmentation for additional qualitative work? There are several ways to avoid the problem, but they all come down to ignoring the "rule" that everyone has to be in a segment and excluding people who do not have a good fit with the segment, screening for just "pure" types.
Yarnell has expertise in both quantative statisical analysis as well as a variety of qualitative methodologies. We are uniquely qualified to work with you to both understand the technical definitions of the segments and to develop statisical-based screening questions that select "pure types" for each segment without the need for a complex Excel spreadsheet. As such, screening for smaller samples used in qualitative research can be more targeted and effective.