Although I dont have a percentage for you, I think your right. With a new launch, you always want to stay conservative with your evaluation. It also depends on what you have launched. Is your product something thats going to generate a buzz? Does the product pre-launch actually work? Regardless, If your creating an evaluation, I would stay on the conservative side.

How long is a piece of string?

The retention rate of beta users is going to vary on such a large number of factors (some of which you mention) as to render comparison futile.

I'd both - agree and disagree with you. :-)

Yes, I guess comparing one case with another would be more or less futile.

However, I do believe some kind of useful/meaningful/actionable intelligence could still be derived from case studies and statistics - don't you think so?

There's a big difference between a case study and statistics.

A case study is just one big, detailed anecdote.

You can learn from it to the degree that your situation matches the situation described-- but you still don't know if the factors that you are matching on are actually the relevant factors or not in terms of the outcome.

Statistics would be wonderful, if there were a way to normalize for all of the unknown factors. Which there isn't, so they aren't.