Because Semphonic is heavily focused on behavioral segmentation, one of the more common questions I get asked is whether or not we use survey and/or customer data when building a segmentation. Yes. Or should I say YES!
I’m a firm believer in behavioral segmentation, but that’s in counterpoise to a world where the vast majority of segmentation is based on survey and relationship data. I believe in behavioral segmentation for exactly the same reason I believe in including survey and customer data; the more data you have, the better segmentation works.
In our two-tiered segmentation model, the first tier is a classic customer-based (visitor type) dimension. It can and should be built using as much data as you have available, particularly relationship data. In the example segmentations I showed, this dimension included things like “Advisor/Administrator/Investor”, potential assets under management, “Travel Agent/Consumer”, “Online Booker/Offline Booker”, “Patient/Caregiver/Health Professional”, etc. All of these are classic traditional segmentation variables that establish the essential relationship between the customer and the company.
On the other hand, what you haven’t seen in my segmentation descriptions are more persona-based or attitudinal descriptions. I haven’t shown a dimension like “Early Adopter/Mid Adopter/Late Adopter” for technology or a dimension like “Brand Promoter, Brand Positive, Brand Neutral, Brand Detractor”. It isn’t that I don’t believe such dimensions are useful. If you have a logged-in or fully-identified digital channel and you have this type of segmentation for your customers, it’s well worth using.
For most digital properties, a significant percentage of behavior is anonymous. This presents a barrier to the full incorporation of persona-based segmentation and to the use of opinion research data. Most surveys touch only a tiny percentage of visitors in a channel. If you build a segmentation using survey data, that means you can’t classify visitors who haven’t taken the survey.
Our goal in creating a Two-Tiered Segmentation is to categorize every Visitor and Visit to the site. Admittedly, this sometimes involves finer degrees of segmentation for known customers, but that reflects the actual use of the data in most real-world cases. In our travel segmentation, we had customer dimensions for the Loyalty Program and Offline/Online Behavior that simply wouldn’t exist for prospects. For the most part, that mirrors the way the business will deal with them. You don’t sell a Loyalty Program before a room and you don’t shift a customer to online unless you have a customer.
Dimensions like “Brand Promoter/Detractor” or “Adopter Type” are just different.
That doesn’t mean there isn’t a profound role for survey data in creating Semphonic’s Two-Tiered Segmentations. There is. In my last post, I showed the iterative process by which a Segmentation is created. Here’s that same diagram with steps colored in red that need or benefit from opinion research data:
In Segment Discovery, the analyst starts by creating a hypothetical segmentation framework. Creating this framework is necessarily subjective and often relies on existing work in usability, on site walk throughs, and on your existing online opinion research.
Nearly every online survey includes two questions critical to the creation and validation of a Two-Tiered Segmentation: Task Intent and Task Accomplishment. Task Intent is designed to capture why the visitor came to the Website. Task Accomplishment is designed to capture the visitor’s subjective view of whether his/her purpose was achieved.
Task Intent is a nearly perfect match to our second tier (Visit Type). I say “nearly” perfect because the match is not quite exact. We often find reasons to shape and change digital use-cases in ways that deepen or go beyond basic customer intent. There’s a good example of this in my Media-Type Segmentation example. I split SEO and Direct-Sourced visitors within a basic informational use-case. It’s a distinction that would certainly not exist in the visitor’s mind and would never show up in a survey around task intent but it exists in the real-world because of the very different success rates of these two cases. With sufficient attitudinal depth, you could pry apart the deeper reasons for the difference, but the behavioral measure is simple, available and powerful and goes deeper than any single Task Intent question is likely too.
Having said all that, the tasks described in your online opinion research are a terrific starting point for describing a 2nd Tier Segmentation and they can be quite a bit more than a starting point. Behavioral analysis by Task Intent (the content and paths most used by this group) can help define the initial set of filters to test.
Where you’ve incorporated your online survey data into your behavioral data, you can also use the survey data for Attitudinal Validation. After you’ve built your filters, you cross-tabulate each segment with the survey data to see how closely your Visit Type filters map to the survey Task Intent. It should be quite close and where the two diverge, you should have compelling reasons why. This mapping of survey Task Intent data to your segmentation filters is also the single most powerful means of validating your approach to the whole organization.
Let’s face it, online data is noisy. The whole point of a Two-Tiered Segmentation is to reduce the amount of noise in your reporting, your analysis and your targeting. The segmentation is not immune from the noise and there is no inherent proof that a Visit-Type segmentation is accurate enough to use. If people in your organization don’t accept the validity of your filtering, you’ll never get if off the ground. The integration of survey data and comparison of Task Intent with Behavioral filtering can provide that validation. Nor is it just a matter of pro forma validation; getting the segment filters right is a challenge. Careful tuning against your opinion research data is a genuine benefit.
After validating the filters, the next step in the process is to further refine the segmentation by exploring the visits that didn’t fall into any of the categories created in the first step (segment discovery). Some of the most interesting segment discoveries come in this step – since these are use-cases that, by definition, weren’t really considered in the creation of the site. Once again, opinion research is extremely useful. What tasks did these visitors say they were engaged in? Why don’t those tasks seem to match obvious behavioral filters? Are there patterns in attitudes or demographics that distinguish this group from well classified use-cases? Your best bet for creating an intelligent classification of these visits is a careful consideration of what makes these visits distinct behaviorally and from an opinion research perspective.
This isn’t the end of the story. As valuable as opinion research is during the process of creating and implementing a Two-Tiered Segmentation, it’s even more important later on. In subsequent posts, I’m going to take up the uses of a Two-Tiered Segmentation for database marketing, targeting, personalization, classic analytics, management reporting, and testing. In almost every case, opinion research data is a key ingredient. It provides the customer perspective on success, the attitudinal, the demographic, and the relationship information that help make a segment “come alive” and lets analysts and marketers alike understand the needs, the concerns and the opportunities presented by that segment.