I did this the slow way over a few months, before I trusted my own conclusion. The thing nobody mentions is that how much of your social graph it wants is the first thing I would look at now. The trend was steady every time regardless of how I framed the profile. Test it cheaply first on balance, and adjust as you go.

I tracked this rather than trusting impressions without rushing it, so take it as tested rather than assumed. Consistently, whether notifications can be tuned turns out to matter far more than the interface.

I went through this same question last year once properly, and the conclusion was not what I wanted. Where it actually landed was how consistent match quality is week to week — that matters far more in smaller places.

Where I would focus your attention:

  • Find out how many profiles in your area have been active this week — it is easy to test and rarely advertised.
  • Whether browsing costs anything — surprisingly hard to find out in advance.
  • Whether a paid feature interrupts the conversation — I would treat this as a dealbreaker.

I reset and started clean once properly, because the advice online contradicts itself. Whichever option you pick, the size of the local pool explains why old recommendations age badly.