Your last six posts averaged 40 likes. The six before that averaged 95. Nothing obvious changed, so you blame the algorithm — which feels satisfying and diagnoses nothing. Engagement drops almost always have a findable cause, and most of them are one of five: format fatigue, posting-time drift, weakening hooks, an inactivity reset, or dying comment velocity. Here is how to tell which one is yours, and what to test week by week.

Step zero: separate a reach problem from a resonance problem

Before touching the five causes, split your numbers. Engagement is a product of two things: how many people saw the post, and what fraction of them acted. Pull your last twelve posts and compute engagement per reach — likes plus comments plus saves, divided by reach or impressions. If reach fell but engagement-per-reach held steady, distribution is the problem, and the causes below about timing, inactivity, and velocity apply. If reach held but the rate fell, the content itself is landing worse — look at format fatigue and hooks. If both fell, start with distribution; reach problems drag the rate down with them.

Also rule out the boring explanations first. Did you change what you post about? Did a holiday period land in your window? Did one outlier inflate the earlier average? Comparing six posts to six posts is a small sample; one lucky post in the old batch can manufacture a fake decline. Remove the single best and worst post from each period and compare again.

Cause one: format fatigue

Audiences habituate. If your last twenty posts are all the same template — the same carousel layout, the same 'unpopular opinion' opener, the same talking-head-with-caption video — your most loyal followers have learned to recognize and skip it before reading a word. The tell: engagement declines steadily rather than dropping off a cliff, and your newest followers engage more than your oldest ones. The check: count how many of your last fifteen posts share one template. Ten or more, and fatigue is a prime suspect.

Cause two: posting-time drift

This one is embarrassingly common. You built your audience posting at 8 a.m., got busy, and drifted to 'whenever I remember' — often late evening. The first hour of engagement seeds a post's distribution, and your first hour now lands when your specific audience is asleep. The tell: a cliff-shaped drop that coincides with a schedule change, while posts published at your old time still perform. Check your last twenty posts' publish times against their reach. If the scatter is wide and the best posts cluster at one time, drift is your answer.

Cause three: your hooks got comfortable

Hook quality decays silently because you stop noticing your own openings. Early on, every first line was fought over; a year in, you open with context-setting — 'I've been thinking a lot lately about onboarding' — because writing feels fluent now. Fluent is the enemy. The tell here is impressions holding steady while everything downstream falls: fewer expands on X, shorter watch time on video, fewer see-more taps on LinkedIn.

Run this test on your last ten posts: read only the first line of each, and ask whether a stranger would have a reason to read line two. 'I've been thinking about onboarding' fails. 'We deleted four onboarding steps and activation went up — here is which four' passes. If more than half fail, your drop is at least partly a hook problem.

Cause four: the inactivity reset

If you went quiet for two or three weeks — launch crunch, holiday, burnout — expect your first posts back to underperform badly, and expect it to persist longer than feels fair. Two things are happening: distribution systems favor accounts with recent engagement history, so you are re-earning your baseline, and your human audience genuinely forgot the thread of what you were doing. The fix is unglamorous: post consistently for two to three weeks at moderate frequency, prioritize the formats that historically earned you comments rather than just likes, and refuse to judge any single post during the recovery window.

Cause five: comment velocity died

Early comments are disproportionately valuable — they signal to the platform that a post deserves wider testing, and they signal to human skimmers that a conversation is happening. If you used to reply to comments within minutes and now reply the next day, you cut your posts' momentum in half without changing the content at all. The tell: your comment counts fell faster than your like counts. Likes measure content; comments measure relationship, and relationships decay first when you disengage.

The week-by-week test plan

Do not fix everything at once, or you will never know what worked. Change one variable per week and hold everything else steady:

  • Week 1 — timing: return every post to your historical best publishing time; change nothing else
  • Week 2 — hooks: rewrite every first line until it makes a specific promise or a specific claim; same formats, same times
  • Week 3 — format: swap your two most-used templates for two you have not touched in months
  • Week 4 — velocity: reply to every comment within 30 minutes for the first two hours after posting

Track one metric per week — reach for the timing week, engagement-per-reach for the hook week, and so on. A scheduling tool or a platform like AI BOSS makes the timing week easy to hold steady, but a spreadsheet and honesty work fine. At the end of the month you will usually find one week that visibly outperformed, and that week is your diagnosis.

The algorithm did not wake up and choose you. Something measurable changed — find it before you redesign everything.

One last honesty check: sometimes the drop is real and none of the five causes apply, because the topic that grew your account stopped being interesting — to the audience, or to you. That is not a diagnostic failure; it is a strategy question. But earn the right to ask it by eliminating the fixable causes first. Most of the time, the fix was a first line and a posting time all along.