If a handful of registrations is enough to skew your results you won't be able to accurately conclude anything from your experiments. This is a good read, the

If a handful of registrations is enough to skew your results you won't be able to accurately conclude anything from your experiments. This is a good read, the stuff about "busy work" - even a bad conversion rate will outperform your best effort at scale, it's very easy to distract yourself with the wrong work when the hard stuff is so much harder. http://blog.samaltman.com/the-post-yc-slump http://blog.samaltman.com/the-post-yc-slump