[jonnathanson]

The problem with the landing-page-and-email-addresses method is that it's impersonal. It's also a low-commitment activity. It's so easy to sign up for something, and it requires so little conscious thought. You will get a lot of false positives because of that.

It's counterintuitive to some extent, but consider a two-phase signup process: a landing page plus an email validation of some kind, or perhaps a two-step signup process on the landing page. There will be a noticeable drop-off between Step 1 and Step 2, and that's good. Those are probably people who would have washed out anyway, and you no longer have to waste time catering to them as false positives. [1]

I'm not saying this was necessarily your issue, but a lot of people run into problems with hypothesis testing by conflating it with marketing or growth hacking. It's not. You're not trying to optimize a hypothesis test for pure signup volume. You're trying to use a hypothesis test as a filter: for the right audience, for your problem/solution statement, and for the value proposition you've chosen to test.

You don't want signup to be a total pain in the ass, but at the same time, you don't want it to be so easy that it loses meaning as a signal of intent.

[1] Caveat being that you still want to probe, and perhaps ask people who've bounced between 1 and 2 why they did so. Sometimes it really is just a UX issue. But in my experience, if someone really has the problem you're addressing, he or she will stick through a two-step signup or validation process.