The 30 minute cookie is a problem The cookie is limited to the session. My guess is that Google will be aware of this and working on some type of a fix if people continue to give their feedback about the test designs and the results their seeing. As I see it, the squeaky wheel gets the grease.

As for the servers…that’s quite a problem. So, if I understand correctly the call to the server ignores the information on the cookie from a previous session? I’ll talk to my friends in the IT department in the morning and see what they have to say about that.

Thanks for reading. Sorry for the delay in the response. I’ll try to speed things up from now on.

Sincerely,

Daniel W. Shields
Analyst/Contributor

1) The default 30-minute cookie timeout meant that we couldn’t automatically limit a user to version a or version b for the entire campaign. We were able to modify the _utimeout variable to address this, but we lost some time before I figured that out.
2) We load balance our site to 2 front-end servers, each with a separate subdomain. Website Optimizer dropped a different cookie for each, which meant that even with the timeout issue addressed, it was impossible to lock them to a or b because they could load balance to the other server when they came back (or at least had a 50% chance of doing so).

Based on that, I don’t think I’ll use it for any sort of large-scale multivariate testing again. It would be fine for testing conversion rates on different ad copy, for example, but we’ll explore other options for anything more comprehensive.

P.S. Could you please get your feed set up so I can stay up to date?