Hold Rate & Retention Curves: How to Read Where Viewers Drop Off

A strong hook buys you three seconds. What happens next decides whether the ad sells. Hold rate and the retention curve are how you see it: who stayed, who left, and the precise moment they left. Read properly, a retention curve is the closest thing paid social has to watching over a viewer's shoulder.

What is hold rate?

Hold rate measures how well a video keeps attention after the opening. Teams calculate it in one of two ways, and it is worth knowing which one a benchmark uses:

On Meta, a ThruPlay is counted when a video is played to completion or for at least 15 seconds, whichever comes first. On TikTok, use 6-second views or average watch time as the closest equivalents and keep the definition consistent between ads.

What is a retention curve?

A retention curve plots the percentage of viewers still watching at each second of the video. It starts at 100% and can only fall: nobody re-joins halfway through. Meta shows a version of it per ad (video retention in the ad's breakdown), TikTok shows audience retention in its reporting, and YouTube shows audience retention for each video.

The value of the curve is its shape. Two ads can share the same hold rate and fail for completely different reasons. The curve tells you which.

Five curve shapes and what they mean

1. The cliff in the first two seconds

Most viewers leave before the hook finishes. The opening is not stopping the right people: it is slow, unclear, or speaks to the wrong problem. Fix the first frame and first line of text before anything else.

2. The ledge, then a cliff

Viewers stay through the hook, then leave together at one specific second. Something happens there that breaks the promise: a jump to a logo, a long product spec, a scene change into a different tone, or the first mention of price. Watch that exact second and cut or move it.

3. The slow, steady slide

No single moment loses people, but the curve never levels off. That usually means pacing: the same shot held too long, a talking head without B-roll, or a script that repeats itself. Tighten the edit and add visual change every two to three seconds.

4. Healthy decay with a flat tail

A sharp early drop that flattens out by the middle of the ad. This is what a good direct-response video looks like: the hook filters, and the people who stay are genuinely interested. Make sure the offer appears while the curve is still high.

5. A late rise in drop-off near the end

Viewers stay through the story, then leave just before the call to action. The ending is too long or signposts "the ad is over". Bring the CTA forward and give a reason to act now.

Where to put the offer and the proof

The simplest use of a retention curve is placement. Find the second at which roughly half of your hooked viewers are gone, and make sure the three things that sell are on screen before it:

1. 1What it is: the product and the outcome, shown, not described.
2. 2Why believe it: a result, a demonstration, a review or a recognizable creator.
3. 3What to do: the offer and the call to action, readable with the sound off.

For many short-form ads, that point arrives well before 15 seconds. A 45-second ad whose price appears at 0:38 is talking to a small fraction of the people you paid to reach.

Eight edits that improve hold rate

1. 1Cut the first beat after the hook. The line that "sets up" the video is often the one that loses people. Go straight from hook to substance.
2. 2Change the picture every two to three seconds. New angle, B-roll, a zoom, on-screen text. Movement resets attention.
3. 3Show, then tell. A demonstration holds better than a description of one.
4. 4Move proof earlier. Results, reviews and before/afters belong in the first third, not the last.
5. 5Use open loops. "The third one surprised me" keeps viewers through the first two.
6. 6Caption everything. Viewers who can read along stay longer with the sound off.
7. 7Shorten. If the curve flattens at a low level by second 20, the remaining seconds are talking to very few people.
8. 8Give the ending a job. Replace a slow outro with the offer and a clear next step.

Benchmarking a retention curve

Compare a curve to two things: your own past winners in the same format, and a category baseline. A curve that sits above the category baseline at every second is a strong ad, even if its absolute hold rate looks low. A curve that tracks the baseline and then falls away at one moment points to an edit, not a new concept.

Common mistakes when reading retention

• Comparing different lengths. A 15-second ad and a 60-second ad will never share a curve. Compare by second, not by percentage of the video.
• Reading too early. A curve built on a few hundred impressions is noisy. Wait for a meaningful sample before acting on a single second.
• Fixing the end first. If 75% of viewers are gone by second three, the ending is not the problem.
• Ignoring the sound-off viewer. Most feeds start muted. If the curve drops where the voiceover carries the message and nothing is written on screen, add captions.