This is one of a series of posts about a real-life attitudinal audience segmentation project. See other posts in this series.
Now we get to the fun part. They survey’s been fielded and the data’s in — now it’s time to get the stats involved and see what segments the data gives us.
This is the part where clients get nervous — especially when they’ve got previous research that has led them to believe they’ve got a handle on their segments. Some have even suggested that we start with existing segment assumptions and use the data to gain more insight on them.
The problem is, that’s not segmentation — that’s profiling. And they’re very different.
There are many approaches to looking at your target population. You can use demographics to group people by gender, or age, or lifestage, or household income. You can group them by the products they buy from you, or those they buy from someone else.
But attitudinal segmentation groups people on how they FEEL about things — which can have little or no correlation to any of those items. Most of us don’t buy something PRIMARILY because of how much money we make, or where we live. Many of us buy the same products as our neighbors or co-workers, although our demographics may be very different.
So if you start an attitudinal segmentation project by pre-establishing demographic or other bases for the segments, you’re missing the point. If you do that then all the work that follows will provide you great insight into the groups you’ve established — but you may completely miss the underpinning reasons that people choose your product or category.
Some statistical approaches to segmentation have the same issue — they may require that you identify a “seed” in the data upon which to base the analysis. Pick a different seed, get a different picture.
Factor analysis, on the other hand, takes all your data and allows respondents to sort of “clump” together based on the answers they’ve given you to attitudinal questions — “how important is this?” or “how interested are you in this?”. There’s simply no external decision-making getting into the middle of the process and mucking it up.
So how does factor analysis work?
Well, first of all you determine which questions on the survey are going to form the basis of your analysis. (In attitudinal segmentation, these will be the questions asking respondents’ opinions.) All of the other questions, while not used for the factor analysis itself, will be used later to gain more insight into each of the segments once they’ve been established by the analysis.
Then you determine how many segments to shoot for. This is based on a few considerations:
1) Are you going to segment the entire respondent pool, or are you interested in determining the segments for predetermined groups (like customers vs. non-customers). I’m a fan of the former — in part because of my issue with pre-segmenting, and also because segmenting everyone together ensures that you’ll identify any commonalities between the groups that you can later use to develop umbrella messaging and broad market strategies.
2) How much data have you got? There’s no point in having eight segments if they’re each too small to provide any statistical reliability. I think it’s important to have confidence in this data for strategic planning, so if you have a small respondent pool to work with I prefer fewer, more reliable segments going with more segments that are less statistically viable.
3) What are you going to do with the segments? This becomes in part an issue of facing reality. How sophisticated is the client — can they get their heads around the nuances between eight segments, or would they be better off with five segments that are more easily differentiated? Are they really going to be able to follow through with marketing strategies and tactics for eight, or would defining their message for four audiences be a more achievable goal? Again, my personal preference is to take the route of fewer segments with most clients, only because it’s their first time segmenting their marketing messaging and tactics in this way and I don’t want them to get overwhelmed — especially when they’re trying to communicate their audience segments to other departments within their organization and get everyone on the same page. No matter which answer they choose, the segments won’t be wrong — but with more segments the differences between them become more granular.
Segment 1: Education-Focused. These respondents feel academic excellence is the most important aspect of choosing a school, and see elements like accreditation, faculty experience and graduation rate as key indicators of excellence.
Segment 2: Location-Focused. This group is primarily looking for a good school that’s not far from home. They are more likely to live near SCU and more likely to have friends and family that have attended the school.
Segment 3: Experience-Focused. These respondents are primarily interested in experiencing traditional residential undergraduate college life. They are more interested in issues like on-campus housing options and extracurricular programs, and actually prefer that a school NOT be close to their home. Unlike the other segments, which included both undergrad and adult respondents, this segment was exclusively found among traditional undergraduate students and prospects.
Segment 4: Career-Focused. This group wants a college that will help them advance in their chosen field. A practical focus and career-skills development are more important to them than other groups, and they’re more like to be found among adult learners than among undergraduates.
Segment 5: Faith-Focused. This segment is interesting in that it probably wouldn’t exist for a non-faith-based school. These respondents are specifically interested in the ability of a school to support their practice of their faith, and a significant percentage of them list the growth of their relationship with Christ as a key reason for attending college.
Each respondent is assigned a number based on which segment they align with most closely. (There’s always some gray area — some respondents will appear to straddle two main segments, in which case we will assign a primary and secondary segment to make future messaging decisions easier.) Then, based on the respondents in each group, it’s a straightforward exercise to evaluate their responses to other items on the survey — demographics, purchase behavior, brand familiarity, etc. — and identify the additional elements which differentiate them from other groups, and those which all groups have in common.
Next: What We Can Do With All This Insight!