just test it (it's very easy to test on sites that get some decent traffic). well, "it's very easy to test" if you are willing to loose some traffic for some ti

just test it (it's very easy to test on sites that get some decent traffic). well, "it's very easy to test" if you are willing to loose some traffic for some time (and as a SEO you MUST be willing to so such tests) also i know the webmaster videos as well, they are more in the tone of "you have to be careful when you look at user behavior". also your argument has huge holes: "it's GOOD if a visitor spends 5 seconds there vs. 5 minutes because it means they found what they wanted." well, there is an statistical technic called "standard deviation", if the SD is too big, don't factor it in. (i don't say that google does this, i'm sure they are more sophisticated, but it's a good counter argument for your example) "Sometimes usage data is misleading." well, then G would just have to determine when it is misleading and with which probability. i repeat: just test it. (and: don't read to much seomoz.)