Segmentation Is Not the Answer

I’m starting to cringe every time I hear the phrase, “We could segment our audience based on that data”.

People hear that phrase and eyes light up and heads bob in agreement. In meetings with the CMO/SVP, the phrase is carefully proffered as a goal about to be fulfilled, and the CMO/SVP solemnly nods and pats everyone on the back for having done a good job.

What utterly useless, well-intentioned bunk.

The reality is, it’s easy to think of many data points on which we could theoretically segment. The challenge is to find data points that would provide value if used. That requires a level of thoughtfulness, and an access to resources, which may not be available. Hence the plethora of talk and the dearth of action and results for many brands.

There are a few questions to ask if you’re going to do segmentation right.

1. Do you have content for that?
   It’s lovely to decide that everyone with blue eyes simply must have content that matches. But do you actually have that content? It’s rare for the email channel to drive its own photoshoots or afford to write whitepapers, articles, or blog posts just for use by the channel. So if you haven’t already created the content for the website or for another purpose, then you probably can’t get the funding to create it just for your emails. Don’t bother segmenting on data that will require you to have new content created if you know the budget and resources won’t be there.
2. Will the difference in content be significant?
   Recently, I reviewed some enterprise-wide segmentation that was done well and looks really smart. It will probably help the client identify best prospects and spend limited sales force resources far more wisely than they could have without it. The differences in content, however, are minimal. The top five items of importance to customers are all the same, just ordered slightly differently. Yet the client believes that there is value in creating different email streams for each segment. Never mind that each email will cover so much of the content that recipients won’t even see the ordering. The difference in content is so insignificant that the whole exercise is pointless. Honestly, there are better things to spend time and resources testing. When the difference in content isn’t minimal, move on.
3. Is your audience big enough?
   Every once in a while, I run into a client who is so excited about segmentation, they micro-segment into an audience of 10 or 100 or so. This is fine if you can automate your content like Amazon.com does, or if your ROI justifies such small segments. But if you are like most marketers and don’t have that level of automation and support, it’s a good idea to size your audience before you extol the virtues of the latest segmentation idea.
4. Is your audience small enough?
   I worked with a client once who broke their audience into groups based on whether they were likely to buy (or looked like anyone likely to buy – we call those “lookalikes”) within a broad category. It didn’t work. The analysis was good, so far as it went, but the client decided to limit the number of categories and group so many “lookalikes” in, that ultimately the groups were too large and amounted to an A/B test rather than true segmentation. In other words, pretty random. So the results were relatively close to what they would have gotten with a generic blast to the entire base. If you’re going to segment, do it right and don’t try to force the data. Create a “doesn’t fit anywhere else” group so you can see the results of your segmentation against the population it is intended to sway.
5. Can you measure your results against a control group?
   Think about how you break your audience into segments, and make sure you have control groups for each segment set up. If you just divide your audience and send different offers against a single control group, you are not measuring the value of your segmentation – you’re measuring whether your data is accurate at best, and maybe not even getting an answer to that question. Segmentation should be a combination of data and creative driving results. So you want to look at both. There are a couple of approaches to this, so make sure your analytics team is involved from the start so that your testing is set up correctly.
6. Is your analytics team on board?
   So many email marketing teams have little support for analytics. But if you want to truly measure segmentation results, you must make sure you have someone smart and savvy from the analytics team on board early in your project. Their input will help you drive counts, understand whether your tests are statistically significant (in other words, the real deal or a fluke that you won’t be able to repeat), and make sure tests are set up correctly so that you can take the great results you get to senior management with confidence.

It’s all well and good to say, “we could get every customer exactly what they need”, but keep some perspective on whether you have budget and resources to justify the effort. Next time someone points out how great it would be to segment on something trivial, use the 6 points above to validate the idea. If it passes, you may have something that will drive fantastic results. If not, you’ve just saved yourself a few cringe-worthy moments.