Market SegmentationMicro TopicsQualitative Research

Market Segmentation, Southwest Airlines Style

The TMRE session was titled, “Segmentation 2.0: Optimizing a Segmentation Model Using a Range of Tools and Stages.” And sorry to be blunt, but “2.0” was misleading.

Or was it?

The session started off benign enough. A classic segmentation study. Start with some qualitative, proceed to quant. SOP.

Key pointers from the session included:

  • Be sure to spend sufficient time planning the project
  • Be sure to have clarity on objectives (how the segmentation model will be used)
  • Include stakeholders in the process
  • Start with qual as a phase 1

All good, basic points, but certainly not 2.0.

But they did do two things not currently done in all segmentation studies.

  1. For the qualitative Phase, Southwest used ethnography. Tammy Sachs was their partner for this phase, and she shared some great video snippets from their ethnographic interviews.  I must say, it was very compelling to hear customers talk about their attitudes and perceptions of Southwest as well as of other airlines.  Those who felt strongly about getting miles—and listening to their passion about it was impressive. Those who value a good deal were also very articulate and compelling. And so on. There is nothing like hearing—and seeing—people  talk ad lib to really get a sense of their attitudes and values.  So ethnography is cool, and applied very well here…but is it “2.0”? Debatable.
  2. They used an “a priori” segmentation model. Yup, that’s right. They went into the study with a hypothesized set of segments in mind. The segments were based on behavioral data from their existing customer database.  During the presentation, this confused me. We were, after all, in a session on conducting segmentation. The process was defined as qual, leading to quant. But the speaker occasionally referred to the segments they started with. Isn’t a segmentation study usually used to derive segments?  Well, not in this case.

Southwest was concerned about having a model that would be actionable with its existing customer database. So they opted to create a segmentation model based on variables they already have, and build from there.  The market research was then designed to do two things:

  1. See if they missed any important segments
  2. Profile the segments they had created from the database

Now I confess, upon hearing this, I was stunned. This is not 2.0 in my mind…this is 1.0.  But after my initial reaction, I digested a bit. And there is some important merit in their approach.

Consider this:

  • They have a model that allows them to easily tag customers into segments (so no risk of having a model that is academically interesting but hard to apply to real business tactics)
  • They have a model that will likely resonate with their decision makers (since it uses variables that are familiar)

So is it 2.0? I don’t think so. But it brazenly defies a lot of current thinking about segmentation. And that is refreshing.

Southwest is often described as a low frills airline that delivers great value. Perhaps this also describes their segmentation approach.

[For more on segmentation, check out this video preview: Video Link]
Tags

Kathryn Korostoff

Kathryn Korostoff is founder and lead instructor at Research Rockstar. Over the past 25 years, she has personally directed more than 600 primary market research projects and published over 100 bylined articles in magazines. She is also a professor at Boston University, where she teaches grad students how to analyze and report quantitative data.

Related Articles

6 thoughts on “Market Segmentation, Southwest Airlines Style”

  1. Great post! I particularly liked your point about how they were able to use the information for actual business action and decision-making, and not just be able to say they conducted an academic exercise. If nothing else, having actionable data will make them more likely to see the benefits of conducting research in the future, which in turn, will maximize their investment and guide them toward even better strategic decision-making. Perhaps at that point, they will really move into 2.0!

  2. Nice wrapup — a depressingly large number of segmentations are never used by the sponsoring organization. It’s something I’ve seen again and again. The academic in me prefers to go off and discover what the research tells us, but the businessperson in me wants a segmentation to be used to help the business grow. I think a hybrid approach such as Southwest outlines is one way to make sure a segmentation is adopted.

  3. I agree w/Jeffrey here. The academic in me is all about starting from scratch but the business realist side of me acknowledges that, if you spend enough time with your existing data and customers you often already have a feel for how these groups might play out. It’s almost like a ‘gut feeling” about how the segments might fall out. I think as long as you don’t let this completely cloud your vision once you’re knee-deep in the segmentation I think it’s ok. It could also serve as a good bet among researchers up front : )

  4. Love your post Kathryn. Interesting case, but definitely not “segmetation 2.0.” I have had the fortune to be involved in segmentation studies that yield actionable results with no a priori model. Those require a lot of work in the design phase so that the resulting data can be operationalized in marketing actions at the end (which also requires that the market researchers have some notion of what marketing is and how it works in the real world). Segmentations that don’t do this end up being useless and collecting dust in some corner. It is all in the design!

  5. I have been involved with a number of projects which made use of apriori segments. Sometimes they work, but when the apriori segments cannot be found or made actionable there are big problems. People have a hard time letting go of a segment they already believe exists.

  6. Hallo Kathryn, very interesting post. Working as market researcher at a German mail order company I have thought several times about using market research in order to find the “best cut” of existing data warehouse dimensions (let’s say 3 dimensions like gender (2 poles), age groups (3 poles) and online yes vs no (2 poles)). That would lead to 2 x 3 x 2 = 12 “cubes” that could be sampeled in MR as quota cells and then further investigated with regard to their best “cutting points” in terms of questionnaire variables like customer needs, attitudes, category preferences etc.. I am confident that this kind of “cube segmentation” could be a good hybrid approach making sure to always have anchor variables within existing data warehouse. Unfortunately, the mail order company ran out of money before I was able to try it but I still think this is a promising concept.

Leave a Reply

Your email address will not be published.