A reply is worth 3x a like as a signal someone actually cared

The stat. In the engagement score likes + 3×replies + 5×bookmarks, a reply is weighted 3x a like. It sits in the middle of the intent ladder: more meaningful than a passive like (it costs effort and attention to type something), less predictive of durable value than a bookmark (which signals you'll come back).

How it was measured. Like the bookmark weight, this is a deliberately-set prior reflecting the cost and intent behind each action, then validated against one account's logged outcomes. Posts that drew genuine replies tended to outperform like-only posts on downstream follows and conversation depth, supporting weighting replies above likes — but bookmark-rich posts still beat reply-rich ones on durable value, keeping replies below bookmarks.

The honest caveat. Replies are the noisiest of the three to weight, because not all replies are positive — a ratio'd hot take racks up replies that are angry, not appreciative. The score treats all replies as equal-value, which overstates the worth of controversy. The 3x is a reasonable middle prior, not a measured coefficient, and a content strategy that optimizes naively for "replies" can drift toward bait. One-account validation, same caveats as the rest of the study.

The tactic it implies. Treat earning a real reply as a higher bar than earning a like — design posts that invite a specific response (a question, a take worth arguing with, a gap someone wants to fill) rather than ones that just feel agreeable. But don't confuse volume of replies with quality: a hundred angry replies is a worse outcome than ten thoughtful ones, even though the raw count looks better. The score rewards conversation; your judgment has to filter for the right conversation.