Bookmarks, not likes, are the strongest signal that a post landed

The stat. The study scores every post and reply with likes + 3×replies + 5×bookmarks. A bookmark is weighted 5x a like and ~1.7x a reply — the heaviest weight in the formula — because bookmarks were the action that best tracked durable value: people save things they intend to come back to, not things they merely reacted to in passing.

How it was measured. This is a modeling choice validated against behavior, not a single observed ratio. Across the logged corpus, posts that accumulated bookmarks were the ones that kept producing downstream engagement and follows, while like-heavy/bookmark-light posts tended to be flash-in-the-pan. Retweets were deliberately excluded from the score — they proved too noisy (driven by a few large accounts) to be a reliable value signal.

The honest caveat. The 5x weight is a deliberate prior, hand-set to reflect intent, then sanity-checked against one account's data — it is not a regression coefficient derived from a controlled experiment. Bookmarks are also harder to see (X hides bookmark counts in many surfaces), so they're a noisier metric to act on day-to-day than likes. Different content types (a reference thread vs. a hot take) will have very different natural bookmark rates.

The tactic it implies. Stop chasing likes as your north-star metric and watch your bookmark rate instead. Likes tell you something felt good for a second; bookmarks tell you it was useful enough to keep. If you want followers and authority, write the thing people save — checklists, frameworks, hard-won specifics — not the thing that earns a reflexive heart and is forgotten by the next scroll.