Overview
For many years, marketing professionals have relied on a set of analysis techniques designed to help them understand the demographic and psychographic profiles of their customers and prospects. These traditional segmentations are usually derived from complex clustering techniques than map rich primary research data (usually survey based) into common groups or profiles. These groups are then given highly descriptive business names and rich descriptions and provide a framework for a wide range of marketing activities. Though such segmentations can be (and are) applied to online customers, companies that have tried to map these segmentations down to the individual level (for targeting or reporting) in the online world have mostly been disappointed. In Part I of this series, I described the biggest pitfall in extending these segmentations – the near impossibility of mapping demographic and psychographic profiles to visitors about whom we typically know nothing except their online behavior. In Part II, I discussed the advantages and disadvantages of building behavioral segmentations. In this post, I’ll discuss adding demographic and psychographic information to the segmentation using Online Survey techniques.
No matter how rich a behavioral segmentation is, it is never a complete marketing answer to the question: “Who are my customers/prospects?” A behavioral segmentation will always fall at least a little bit short since it does not capture any demographic categories and it can provide only inferences about the attitudes driving actual behavior. Both of these are important short-comings.
Demographic categories are big part of how we think of people. We may know that someone is a sports enthusiast, but our understanding of who they are is still colored by their age and gender. We think about a 27 year old female sports enthusiast as completely different from a 82 year old male or a 9 year old boy. We may know that someone is interested in private aviation and flying planes, but how much attention they get in a lead assessment may still be powerfully determined by their income. No type of information has proven as consistently interesting, useful and actionable as demographic categories. So demographic categories definitely matter. The same is true – if not equally so – of the attitudinal variables that also drive traditional customer segmentations.
Although attitudinal questions can be the most powerful and illuminating variables in a traditional segmentation (and certainly the most business specific) they have always suffered from the fact that they cannot always be translated into any other marketing purposes. As targeting variables, they simply don’t exist outside the segmentation; so marketers often have to find behavioral or demographic proxies for these variables anyway. That being said, no proxy for satisfaction, interest, commitment or brand perception will ever remotely match the actual words and answers of consumers when specifically asked about these things.
Depending on your situation, there are a couple of different approaches to adding demographic and attitudinal information to your segmentation.
- For customer segmentations, you can join demographic and business relationship data directly to the web behavioral stream of secure/registered/partly-secure sites and include these variables directly in the segmentation.
- For customer segmentations on non-trackable sites and for most prospect segmentations, you can join demographic and psychographic data from online survey research directly to the web behavioral stream and use the data directly in your segmentation OR
- You can join demographic and psychographic data from online survey research to the web behavioral stream AFTER the segmentation and use the data to color your segmentation.
For situation #1 (customer data on secure or trackable sites), it would foolish not to add every important piece of customer data to the behavioral segmentation. This is data you have about every customer (so it doesn’t impact your ability to do targeting) and it can provide dramatically improved color and context around the behavioral segmentation. Key customer variables are different for every business and industry, but you’ll almost always use basic demographics and geographics, length, size, number and type of customer relationships, cost and profitability data as available, and other touchpoint data at some summarized level.
For situation #2, you will run the online survey first and then build the segmentation from the combined web behavioral and survey data. Typically, you’ll include a survey identifier in the web behavioral stream as the survey runs. For Omniture, there are plug-ins that allow you do this pretty seamlessly for the most popular survey solutions and it is pretty trivial even if you are using a home-baked solution. This process should get even easier if you end up using Omniture’s new survey solution (which I think looks pretty exciting). If you have a compact survey with few questions, you may choose to simply capture survey responses in real-time and move them into custom variables. In other cases, you’ll match back completed survey information using the survey identification.
For situation #3, you will usually run the online survey during the time period you wish to study, but it can run after the basic segmentation and segment classifications can even be passed to the survey. In most cases, however, you'll use exactly the same integration method as in #2.
The choice between #2 and #3 is by no means a slam-dunk. If you bake the online survey data into the segmentation, you’ll probably produce better (by which I mean crisper and more interesting) segments. But you’ll lose the ability to replicate them for any online targeting, reporting or analytic purposes. That’s a lot to give up.
So I generally favor using the demographic and psychographic segmentation to color a prospect or non-trackable site segmentation.
This “coloring” process is relatively straightforward. After you’ve matched back the survey data to an online visitor, you’ll have a data set that includes the visitor id, the survey id, the survey responses, and the behavioral segment you assigned that visitor to.
Using this data-set, you can profile each behavioral segment using the demographic and psychographic variables. Typically, this involves producing both a segment distribution and segment average for each online survey variable. Once you have these descriptives, you can quickly determine where the behavioral segments exhibit strong differences from the site average for any of the survey variables.
Even though the survey data isn’t used to build the segments, it can help describe them. And that, really, is the point of the coloring. If you’ve identified a behavioral segment and it turns out that it skews young and male, then this can become a significant and useful part of the segmentation descriptions.
Similarly, if you find that a behavioral segment tends to do very poorly in terms of visitor propensity to recommend a site, this becomes a critical part of the behavioral segment’s profile.
This illustrates one somewhat unexplored advantage to behavioral segmentations - they are quite powerful analytically. If you started with the survey data and tried to find the behaviors that were strongly associated with dissatisfaction, there's a pretty good chance you couldn't do it inside a web analytic tool. Comprehensive measures like satisfaction will rarely tie to a single behavior. So even though the behavioral segmentation isn't specific to a particular analytic problem (like causes of dissatisfaction), it may nevertheless yield powerful insight into them.
I mentioned in an earlier post that there seems to be little logical difference between building a segmentation based on demographics and psychographics and coloring it with behavioral data versus building a behavioral segmentation and coloring it with demographics and psychographics.
But even though these two approaches look like they should produce similar results, they don’t really seem to. It’s my belief that much of the reason for this is that the behavioral data is preferable for clustered segmentation (there are more variables, the variables are not categorical and have more intrinsic numeric weighting, and the groupings are harder to develop without machine assistance) and the survey data preferable for coloring (easier to understand, simpler categories, less numeric weighting, more directly accessible by the analyst).
And there is, of course, the one big practical difference I keep repeating. The coloring data is insufficient to predict segment membership. So while you can color traditional segments with behavioral data, you can’t assign visitors on your web site to segments unless your segmentation is behavioral. That, ultimately, is the critical fact that makes behavioral segmentation more useful.
In my next post, I’m going to take a deeper look at what’s involved in creating a behavioral segmentation from online data.
Comments