[leahculver]

Not at all! I think it's totally valid to use some gut instinct here.

My instinct is that there's a scale from Facebook -> Twitter -> GitHub in the services that our target audience uses (less nerdy -> more nerdy).

Since GitHub performed better than Twitter (by a significant amount, although that's debatable) then I feel that the Facebook end of that scale isn't worth even testing at all. Instead, it would be more useful to try services on the GitHub end of the scale, such as OpenID or maybe something more novel like BitBucket, Heroku, Dribbble etc.

[henrikberggren]

I get what you mean and it's probably a fair assumption. But, the data you showed and the conclusion you draw "Won't be adding that FB button soon" is totally illogical. You have no idea how many ppl. that are coming to grove, want's to sign in but don't because they have neither a twitter or github account and/or don't want to sign up with it.

You need to add a FB signup button, test the increase (or non increase) of signups and then draw conclusions. Or measure bounce rates in some smart way. Your current method isn't very scientific.