How to read your TikTok retention graph — the four drop shapes and what each one means

No API exposes retention curves — not to creators, not to analytics tools, not to anyone. Which is why most advice about "hooks" and "pacing" floats free of evidence. But the curve itself is sitting in your own analytics, and it's the single most decisive chart your account produces: it names the exact second people gave up.

This is how to read it.

Where to find it

• TikTok: TikTok Studio → Analytics → pick a video → scroll to viewer retention. You get the full curve: percentage of viewers still watching at every point of the video.
• Instagram: open a reel's insights. You get average watch time and, for some accounts, a retention graph on the professional dashboard. Less granular than TikTok's, but the average-watch-time number still supports the ratio checks below.

Screenshot the curve for your last several posts — best and worst. The comparison, not the single chart, is where the information is.

The number to check before anything else

How much of the audience is still there at second 3. That's the hook hold. Everything else about the video — the editing, the payoff, the CTA — is irrelevant to the people who already left. If your worst posts bleed out a large share of viewers in the first 3 seconds and your best posts don't, you already have your diagnosis, and it's the cheapest one to fix: it's about the first sentence and the first frame, not your whole style.

The four shapes

1. The cliff (steep drop at 0–3s)

The curve falls off a ledge immediately and then flattens. The people who survived the open actually liked the video.

This is a hook problem, and only a hook problem. Common causes: an opening that describes what the video will be instead of starting it ("so today I'm going to show you…"), a first frame with nothing to read or look at, or a hook that promises a different audience than the video serves.

Fix: rewrite opens to start mid-action, put the claim or the tension in the first line, and make the first frame carry text or motion. Test against your next posts' 3-second hold.

2. The bleed (steady downhill slope, no flat sections)

No single moment kills it; viewers leak out continuously. The video never gives anyone a reason to stay for the next ten seconds.

This is pacing and structure: sections that run long, no open loop ("wait for the third one"), the payoff visible from a mile away.

Fix: cut ruthlessly — most bleeding videos are simply too long for what they deliver. Add a structural promise early and pay it off late. Your duration-band stats (which lengths your account actually holds) matter more than any platform-wide rule.

3. The step (flat, then a sharp drop mid-video)

Retention holds, then a chunk of the audience leaves at the same moment. That timestamp is a specific decision you made: usually the pivot to selling, a topic change, or "before we continue, follow me."

Fix: go to the exact second and watch what you did. Move the promo past the value, or cut the mid-roll ask entirely. The step will tell you next week whether you were right.

4. The plateau with an end drop

Strong hold, then everyone leaves in the final seconds. Not actually a problem if the drop is at the natural end — but if your videos rely on looping (short clips that replay), an end-drop shape can mean your loop point is visible and viewers bail instead of rewatching.

Fix: trim the outro to nothing. End on the payoff frame, not a sign-off.

Two ratios worth writing down

• Hook hold = % of viewers remaining at ~3 seconds. Compare across your own posts; the absolute number varies by account and niche, so your best-vs-worst gap is the signal.
• Average watch ÷ duration. A 20-second video watched for 12 seconds (60%) is holding better than a 60-second video watched for 20 (33%) — even though the second one has more watch time. This ratio is the honest way to compare videos of different lengths.

Make it a habit, not a one-off

One curve is trivia. The comparison across your top five and bottom five posts is a diagnosis — the shapes repeat, and the repeating shape is the habit. Screenshot the curves as you post, keep them with the spreadsheet from the baseline method, and re-check after each week's change. When the 3-second hold on your new posts moves, you'll know the hook fix was real — measured, not vibes.