The short answer
Hold rate is the share of 3-second viewers who keep watching, usually 15-second plays (or ThruPlays) divided by 3-second plays. Practitioners put the average around 40% to 50%, with 60%+ called strong (directional). On our board, video ads that name the brand early in the script reach the top tier 11.5% of the time, against 8.9% for those that don't.
Hold rate in one line
Hold rate tells you how many of the people your hook stopped stayed to hear the rest:
Hold rate = 15-second video plays (or ThruPlays) ÷ 3-second video plays × 100
Hook rate measures the stop. Hold rate measures what the next twelve seconds did with it. Neither is an official Meta metric. Both are custom metrics built from Meta's standard video columns, and teams don't always use the same numerator, so check the formula before you compare numbers across reports. (Motion uses 15-second plays. Vaizle uses ThruPlays.)
Directional benchmarks
| Source | Published hold rate range |
|---|---|
| Motion | Average 40% to 50%, over 60% strong |
| Vaizle | Aim for about 40% to 50%, depending on length and audience |
Both come from agency and tool-vendor experience, not from Meta, so use them as a sanity check. If your hold rate sits under 30% on a 30-second ad, the body is losing people. If it sits above 60%, the body is doing its job and your problem, if you have one, is somewhere else.
What we can see without anyone's view data
Hold rate lives inside each advertiser's account. The Meta Ad Library doesn't publish it. What we have instead is the structure of 1,531 winning video ads on the Ad Radar board with a transcript long enough to read past the opener, plus how long each one has run, its estimated spend and whether it reaches the top tier (top 10% of the board by winner score, about 10% of ads).
An ad that runs 200 days with six figures of estimated spend has a body that holds enough people to pay for itself. The structure of those bodies is what you can copy.
Length: where the top tier sits
| Video length | Ads | Reach top tier | Median est. spend |
|---|---|---|---|
| 15 to 30 s | 183 | 8.7% | $215K |
| 30 to 60 s | 618 | 9.9% | $255K |
| 60 to 120 s | 483 | 10.6% | $248K |
| 2 to 5 min | 421 | 8.3% | $227K |
| Over 5 min | 105 | 5.7% | $191K |
The sweet spot on the board is 30 seconds to two minutes. That is long enough that hold rate matters a great deal (anyone who leaves at second 15 never sees the offer) and short enough that a body built for holding can reach the end. Past five minutes, the top-tier rate drops to 5.7%. More on length in best video ad length.
Four structures that hold
We read the first lines after the opener on the longest-running winners. Four structures repeat.
The countdown you've been promised
The opener promises a list, the second or third line states its size, and the body works through it. The viewer knows how much is left, which is a reason to stay.
This is how your foot was born.
- Est. spend
- $1.5M
- Days live
- 118
- Format
- Video, 293s
- Variants
- 2
Line two: "This is what 40 years of shoes did to it." Line three: "Let me show you which four shoes are the worst." By second ten the viewer has a picture, a stake and a countdown.
The timeline
The script moves through time in fixed steps, so each line is a small payoff and the next step is already promised.
Years of plasma science made in America, USA.
- Est. spend
- $2.7M
- Days live
- 133
- Format
- Video, 207s
Lines four to six: "Week one, I didn't see a thing. Week two, he met me at the door. Week three, climbed up on the bed." Admitting week one did nothing makes week two believable.
The comparison that turns
A question sets two sides, and the body takes one side immediately, with a "sadly" or a "meanwhile" that tells the viewer the story has a turn coming.
Who has a better life, cats or dogs?
- Est. spend
- $9M
- Days live
- 567
- Format
- Video, 124s
- Variants
- 6
Over 18 months live with six variants. The second line starts with "Sadly", and the body contrasts a dog's day with a cat's before the product appears.
The short-sentence story
Stories hold when each sentence is a new fact. Black Skincare Hacks (est. $1.91M, 128 days, 185 seconds) opens "I ran into my ex-husband's new wife at Walmart." and follows with "She was 34." Then "She looked at me like she felt sorry for me." Then "That was in March." Four sentences, four new details, no adjectives doing the work. See story hooks for more.
Two script signals in the data
We checked a few measurable traits of the first ~600 characters of each script (roughly the first 30 to 40 seconds of speech) against top-tier rate.
| Script trait | Ads | Reach top tier | Median est. spend |
|---|---|---|---|
| Brand named in the first ~150 characters | 191 | 11.5% | $244K |
| Brand named later in the first ~600 | 217 | 11.5% | $289K |
| Brand not named in the first ~600 | 1,123 | 8.9% | $221K |
| A turn word ("but", "so", "then", "until") opens line 2, 3 or 4 | 214 | 12.1% | $236K |
The differences are modest, and a correlation in a board of winners doesn't prove cause. Two readings are still useful.
First, ads that name the product early hold up better than ads that save it for the end. A viewer who knows what is being sold by second 30 and keeps watching is a viewer who might buy. One who leaves before a late reveal never finds out what the ad was for. The ads with the highest median estimated spend name the brand after a short setup, not in the first sentence and not at the very end.
Second, scripts that turn early ("But...", "So...", "Until...") reach the top tier more often. A turn is a promise that what follows is different from what came before, which is the job of the body.
How to raise your hold rate
- Find the drop. In Ads Manager, look at video plays at 25%, 50%, 75% and 95% next to 3-second plays. The steepest drop marks the line to rewrite.
- Pay off the hook by second 10. If the opener asks a question or names a threat, the body must start answering within two or three sentences. A hook that waits too long for its payoff produces exactly the high hook rate, low hold rate pattern.
- Promise a structure. "Four shoes", "week one, week two", "two things". A visible shape tells the viewer how long they're signing up for.
- Name the product before the halfway point. On our board, ads that never name the brand in the first 600 characters do worst.
- Cut the second sentence first. It is the most common place for filler ("So today I want to talk about..."). Replace it with a fact.
- Keep long ads for long stories. Over five minutes, the board's top-tier rate drops to 5.7%. Long ads can win, but they need a structure that justifies every minute.
Hook rate and hold rate belong on the same dashboard. Read hook rate benchmarks for the first number and creative KPIs for how both fit with cost per purchase. If you want to read the full script structure of long-running competitor ads, the Ad Radar board lets you search every transcript.
Figures marked as estimated spend come from Ad Radar's model of engagement on public Meta Ad Library ads. They are estimates, labeled as such, and are best used to rank ads against each other.