The personal blog and forum of Max Kaye.
Background: we (Nathan and I) built a political party in ~3 weeks on FB via ads. Cost about $3500 or so.
The biggest thing I learnt was to try as many things as possible via split tests, and if possible try them quickly. We used https://adespresso.com and I can't recommend it highly enough. If you're going to spend more than a few hundred dollars it's well worth it (basic is $50/month and it's not on a contract or anything - maybe even has a free trial).
Particularly a few words can make a HUGE difference (like 30%+ clickthrough rates). An example of that was the following two:
Representative Democracy has failed us
Representative Democracy has failed us, time to do something about it
Particularly I'd suggest split testing images and headlines, since they're the bits that grab people. We noticed a smaller difference with the descriptions/wordy bits, but still significant (15% or so). If you can nail both then there's an easy 50%+ you can just iterate towards, which goes a long way esp. with a NFP or when you're on a budget. Images also can make a 20%+ difference, so make sure to try a few.
Also, make sure to install the facebook tracking pixels so you can track conversions accurately. Facebook can tell you a lot about the people clicking through and signing up just by using that. For example, most clickthroughs we had were older folk (70% male, 50% over 50 or so, which was surprising for a tech based political party) HOWEVER, most signups were younger folk, so you can sort of start to tailor things better with that data.
I wouldn't recommend split testing over interests because you end up with smaller running ads, and you get smaller sample sizes => more money to get significant results. (And you also get info on who is interested in what anyway)
You get a frequency measurement through adespresso too; keep an eye on that because if it stays low you can pump the ad way more (as the same ppl aren't seeing it again). As frequency gets higher your dollars won't stretch as far.
I'd suggest heavy experimentation early on and then pump the ads that work really hard later on. That way you build engagement.
Oh yeah on that point, split tests create multiple ads, that means that comments are only on one single ad, not all of them. Engagement matters so be sure to try and not create ads later in the campaign because you lose that engagement.