Why we rank posts by the creator's own baseline, not raw views

Most "find viral content" tools have the same blind spot: they surface whatever got the most views. That mostly rediscovers big accounts.

A 500k-view post from someone with 2M followers isn't a signal — it's just Tuesday.

What we care about is the post that beat its own author.

Someone whose posts usually do 20k suddenly does 600k.

That 30x gap is the interesting number, because the audience, the niche and the posting cadence are all held constant.

The only thing that changed is the post itself.

So Sona ranks by multiple-of-baseline rather than absolute reach.

It means small accounts show up next to huge ones, and it means a 12x from a 5k-follower writer is treated as more instructive than a 200k-view post from an account that always does 200k.

Two things I'm still unsure about, and would genuinely like other opinions on:

1. Baseline window. Trailing 30 days reacts fast but gets skewed by one prior hit. 90 days is stabler but slow to catch an account that's genuinely levelling up.

Right now I use 30 and I'm not confident it's right.

2. Floor. Below some follower count the multiples get silly — 50 views to 3,000 is a 60x but it's noise. I clamp it, but every threshold I pick feels arbitrary.

If you've built anything that ranks content by performance, how did you handle the baseline problem?