How the engagement score works
and an honest answer to “is this real, or did you just make it up?”
The formula
WER = (6·comments + 3·reposts + 1·reactions) / followers × 100
Posts are ranked by WER (weighted engagement rate). Median WER is just the middle value across the loaded posts — so one viral post doesn't distort it. Only posts where we know the author's follower count get a score.
Why it's built this way
- Divide by followers — we measure an engagement rate, not raw likes. Engagement rate structurally falls as an account grows (small accounts run ~6–10%, 100k+ accounts ~1–3%), so ranking on raw counts just rewards whoever has the biggest audience.
- Comments > reposts > reactions — LinkedIn's 2026 algorithm rewards comments far more than likes (published analyses put a comment at ~15× the weight of a like; longer comments more still), and a repost is higher-intent than a tap. So a comment counts 6×, a repost 3×, a reaction 1×.
- Median, not average — one viral post shouldn't define a creator's typical performance; the median ignores the outlier. It's why analytics tools like AuthoredUp report median.
- Compare within follower tiers — a 5k-follower account will always out-rate a 500k one, so we bucket by tier and rank inside each.
The honest part
The four principles above are each backed by 2026 LinkedIn engagement research — they aren't guesses.
The exact weights (6 / 3 / 1) are a deliberate heuristic I chose to encode “comments matter most” — a compressed version of that ~15× figure (compressed so a single comment doesn't swamp everything). They are not fitted to LinkedIn's real ranking model — that model is proprietary and leans on signals no public data contains (dwell time, first-hour velocity, topic relevance). No public metric can reproduce it.
So treat WER as a compass, not a GPS: it's for relative comparison within your peer set — “which posts and patterns punch above their follower weight?” — not an absolute or predictive number. It is a more honest ranking than raw likes, and good enough to spot what's working.
Want it grounded harder? Once we have real reach/impression data on Kalpit's own posts, we can calibrate these weights to what actually correlates with reach — turning the heuristic into a fitted model. The weights are one line (WEIGHTS) in the code, trivially changeable.