Why are my Reels views dropping? Run the autopsy on your own last 30 posts

Your views dropped and nobody will tell you why. Instagram won't. The comments section won't. And the person on YouTube yelling about "the new algorithm change" definitely won't, because they don't have access to the only dataset that matters: your account.

Here's the uncomfortable part. For most accounts the drop isn't the algorithm punishing you. It's a habit — something you started doing, or stopped doing, across your recent posts. Habits leave fingerprints in your numbers. This post shows you how to lift them.

The three reflexes that waste a month

When reach falls, almost everyone does one of these:

• Blame the algorithm and wait. Unfalsifiable, unactionable. Even when a distribution change is real, you can't act on it — you can only act on your inputs.
• Copy whatever went viral this week. Someone else's winner worked because of their baseline, their audience contract, their delivery. Copying the format transplants the organ without the blood supply.
• Post more. Volume amplifies whatever you're already doing. If a suppressing habit lives in your content, posting more runs the habit more often.

All three share the same flaw: they don't look at the one place the answer actually lives — the difference between your own posts that worked and your own posts that didn't.

Your account is a dataset

Any account that has posted ~30 times in the last 90 days has enough data for a real diagnosis. Not "guru opinion" — arithmetic.

The core idea: your baseline is the median views of your recent posts. Every post is then a multiplier against that baseline. A post at 2.1x did something right. A post at 0.3x did something wrong. One outlier means nothing; a cluster of 0.3x posts that share an attribute is a diagnosis.

Here's the manual version. One spreadsheet, one evening.

Step 1 — build the table

One row per post, last 30 posts. Columns:

1. Date posted
2. Views
3. Duration in seconds
4. The hook — the exact words/text shown in the first 3 seconds
5. Topic, in 2–5 words ("gym fails", "client red flags", "meal prep")
6. Format (talking head, voiceover, text-on-screen, skit, tutorial)
7. CTA type (follow, comment, DM, link, save/share, none)

Getting the hook verbatim matters. "I kind of remember it" is where diagnoses go to die — rewatch the first 3 seconds of each post.

Step 2 — compute the baseline

Take the median of the views column, not the average. One lucky viral post inflates an average and poisons every comparison; the median ignores it. That single number is your 1.0x line.

Step 3 — score every post

Add a column: views ÷ median. Now every post is a multiplier. Sort by it once and just read the top five and bottom five. Most people spot the first pattern before doing any more math.

Step 4 — bucket and compare

Group the multipliers by each attribute — topic, format, duration band (<20s, 20–40s, 40–60s, >60s), CTA, hook style — and take the median multiplier of each bucket.

One honesty rule, and it's non-negotiable: ignore any bucket with fewer than 3 posts. Two data points is an anecdote. Extrapolating from n=2 is how people convince themselves Tuesday at 9am is magic.

Step 5 — read the gaps

You're looking for the spread between buckets on the same attribute. When "tutorial" sits at 1.4x and "vlog" sits at 0.5x across enough posts, that's not a mood — that's your audience telling you what they subscribed for.

The suppressors worth checking first

Patterns that show up over and over when accounts run this exercise:

• Throat-clearing opens. "Hey guys, so today I wanted to talk about…" — the content starts at second six. Compare the multipliers of posts that open mid-action versus posts that open with a wind-up.
• Self-promo in the first half. Pitching your product, page, or course before the value lands. Bucket posts by "promo before the midpoint: yes/no" and compare.
• Topic drift. Your proven lane performs; the experiments outside it don't — but you keep splitting your calendar 50/50. The buckets make the cost visible.
• Duration mismatch. Some accounts hold attention for 60 seconds; most don't. Your duration-band buckets settle it with your numbers, not a platform myth.

The retention cross-check

Views tell you that a post underperformed; retention tells you where it died. Open your analytics (TikTok Studio shows a full retention curve; Instagram shows average watch time) for your best 3 and worst 3 posts and compare the first three seconds. If your worst posts consistently lose a big slice of viewers by second 3, you don't have an algorithm problem — you have a hook problem, and it's fixable this week.

Change one thing. Then actually verify it.

The diagnosis ends with a single instruction, not a rebrand. Pick the worst pattern with the most evidence and change only that for one week. Same cadence, same topics, same everything else.

A week later, recompute: median views of the new posts ÷ your old baseline.

• Meaningfully above your baseline across several posts — the change is probably real. Keep it.
• Meaningfully below — revert it.
• Inside the noise band, or fewer than 3 posts — inconclusive, and that's a legitimate verdict. Run it another week instead of declaring victory.

If you change four things at once and views recover, you've learned nothing — you can't tell which change did it, and you're now superstitious about all four. One variable per week is slower and it's the only version that compounds.

That's the whole method: baseline, multipliers, buckets, one change, verify. It costs an evening in a spreadsheet, and it beats a month of guessing — because it's the only analysis that runs on the account you actually have.