Normalize engagement rate so tiers compare fairly

Normalize engagement rate so tiers compare fairly

Raw ER punishes large accounts. Fix the math:
— Use ER-by-reach, not ER-by-follower, where reach is available
— Compare within tier: nano median 4-7%, macro median 1-2%
— Strip giveaway posts, they distort the median up
— Weight saves and shares 2x likes for buying intent
— Window it: trailing 12 posts, drop the top and bottom

Comparing a nano's 6% to a macro's 1.5% as if equal is a category error.

So what: rank creators against their tier median, never against each other.

Benchmark: tier-normalized ER predicts conversion better than raw ER by ~1.4x (n=200 posts).