Tim Ash says there are many common misuses of statistics for landing pages. It's important to test and filter the traffic, he writes, suggesting the removal of "unstable sources," such as nonrecurring
email traffic, and some "larger but highly variable affiliates" from your testing mix.
Beware of bias samples. You can even run into biased sample issues by running tests that split the
available traffic that shows different versions of the site design in parallel. Ash explains that experiments that include high-data collection rates may be especially prone to this problem..
Read the whole story at Search Engine Watch »