Matt Bailey of SiteLogic did a nice presentation on segmentation at the “Web Analytics & Measuring Success” session on Tuesday morning at the SES San Jose 2007 show. He used an example based on the original Star Trek to illustrate it.
The example began with the belief that crew members who wear red shirts appear to have a much greater likelihood of getting killed when they beam down to the planet with Captain Kirk. So Matt decided to dig into this further, and dug up the following stats:
- 72 percent of the crew members who died were wearing red shirts.
- 57 percent of the time, when a red shirt beamed down to the planet with Captain Kirk, s/he died.
- However, when Captain Kirk hooks up with a woman on the show, the red shirt survival rate goes up to 80 percent.
- Since this appears to be a major survival factor, the next question becomes, how often does Captain Kirk hook up? The answer is 30 percent.
This is a great example of basic segmentation analysis. The analysis began with a question: “Do crew members wearing red shirts have a greater chance of dying when they beam down to the planet with Captain Kirk?” This is already a segmentation step because crew members wearing red shirts are only a portion of the crew (there are crew members wearing other colored shirts as well). Note, to relate this to your Web site, it is helpful if you can think of a crew member dying as a “conversion.” (Matt’s saying this in the session caused quite a bit of laughter in the audience.)
But then, the analysis goes further, breaking down the fate of the red shirts in more detail. Once you have determined that one customer segment converts better than others, it often pays to dig in a little further. In the example provided, whether or not Captain Kirk hooks up with a woman on the show is an additional segmentation step. What we saw there was that the conversion rate went down (crew member survival rate went up).
Isolating Customer Behavior
In a real world Web site situation, this type of segmentation analysis can help you isolate the behavior of different groups of site visitors. Of course, you don’t know what shirt color they are wearing, but there are many other factors you can determine. Here are some examples:
- Does custom behavior vary based on their location? For example, are European visitors more likely to buy red widgets, and are North American visitors more likely to buy blue widgets? If you had this type of information, you could tweak your site to segment the response catering to the differences in the respective visitors by location.
- Does customer behavior vary by source? In other words, which groups provide the best response: organic search engine visitors, PPC search engine visitors, or visitors referred from other sites? Of course, you can break this down further and look at individual search engines, or individual referrers.
- How does a visit by a past customer differ from other visitors? Can you provide a custom experience for a past customer that will further enhance your results?
- You can also compare the behavior of first time visitors with the behavior of repeat visitors.
- Another clear opportunity for segmentation is when you have people who are “members” of your site. It is quite likely that their behavior will be different from that of non-members.
- How does the behavior of people who landed initially on your individual product pages compare with those who landed initially on your review article pages?
Using Segmentation Data
One really interesting way to use segmentation data is in combination with your Site Overlay report (some analytics packages call this the Navigation Report). If you have two or more segments defined, it is very powerful to see how visitor click-through behavior varies by visitor segment in a visual report, and this is exactly what this gives you.
Segmentation is one of the most powerful tools that analytics packages provide. Once you segment your data, you start to get real insight into what is happening on your site. It is really difficult on most sites to get much insight by looking at the cumulative behavior of all your visitors as one group.
Taking Action on Your Site
Once you have segmented the behavior of your visitors on your site, you need to be prepared to take action. Now that you know European visitors buy more red widgets, what can you do to customize their experience by making the red widgets easier to find? Are organic search engine visitors less likely to buy because they are really still in research mode? If this is the case, can you help them find the research information faster, and hence do a better job of setting them up to buy from you in the future?
The questions vary by site. To get the best results, start by brainstorming what those questions might be. Focus on questions whose answers might cause you to do something differently with your site. The best results are obtained when you have the ability to treat various customer segments differently; but even if you can provide only one site experience for all visitors, you may still be able to make other changes to your site (such as site layout) to take advantage of the information you uncover once you segment your visitors.