I started sceptical without rushing it, which is why my answer is not the popular one. For me, how support responds to a real problem accounts for most of the variance I saw. It became obvious without exception even when I varied the time of day.

I took the unglamorous route without rushing it, and the conclusion was not what I wanted. Consistently, the actual reply rate changes noticeably month to month. It held steady again and again across every option I tried.

I gave each one a full month a while back, so the sample is at least reasonable. The short answer is that whether read status is shown consistently accounts for most of the variance I saw.

What I would tell someone starting from scratch:

  • Whether prompts give people something specific to answer — it matters more after the first month.
  • Ask how long before the first upsell prompt appears — worth two minutes before you commit.
  • Whether the sign-up asks for a phone number — it matters more after the first month.

I read a lot of conflicting advice first before deciding, which is why my answer is not the popular one. Boiled down, how wide the distance filter goes is the part the marketing never mentions.