First Level Segmentation
The battle to extract value from ANY segmentation exercise comes down to how well, and how quickly segments can be identified that respond to different treatment. There are several parts to this. First segments need to be selected. These segments need to be big enough to matter, and each segment needs to react consistently in an exploitable way. For example, what If we cut all of our visitors into only two segments; those who we know have visited the site, because we can see our cookie on their browser, and those who do not have a cookie.
This grouping should be useful. If our entire traffic can be split in only two parts, each part should have enough traffic to create a useful sized population. Second, it is reasonable to believe that people, who have been to the site once, and recently enough so that the cookie still exists, probably have some useful characteristics. For example, if someone has seen the site, they might not need as much "about us" and "privacy and security" content.
If we select segments that have too few members, it will be hard to execute tests to determine their preference since each test may take 4 weeks or longer, and it may take 3-4 tests to figure out what elements matter and to find an optimal version of these elements.
If the segment seems to have continuity, but does not behave in an exploitably different way than the segmentation is no good. For example, people who use Netscape vs. Explorer as a browser may provide us with a big enough population, but belonging to either of these groups may not supply us with actionable difference.
Finally, segments that take too long to identify make it very difficult to amass enough of a population to perform a test. For example, visitors who have come three times and reviewed all of the relevant content about a product or service are a very strong segment as far as similarity of purpose and purchase readiness, but the number of people who match these criteria again may be too small to exploit.
So where to start? Try segmenting by known vs. unknown, paid vs. natural, visit 1 vs more than one, customers vs. non customers, and geography. Start with simple changes. Ask if there is any content that can be REMOVED to make the site more relevant to the segment. Try to change one element on each of the top three pages of your most common click paths to increase relevance to the segment.
