What is Hold rate?

A video records 62,400 three-second plays and 7,488 ThruPlays, Meta's count of a video watched to the end or for at least fifteen seconds. Hold rate is 7,488 ÷ 62,400 = 12%. Of the people the opening kept, twelve percent were still watching twelve seconds later. Against 480,000 impressions the same 7,488 is 1.6%, which is the figure that describes the campaign; the 12% is the figure that describes the edit.

Hold rate divides a longer-threshold view count by a shorter one. The denominator is not impressions; it is the people who already survived the opening.

Hold rate = fifteen-second views ÷ three-second views

That choice of denominator is the whole design. Hook rate measures whether an opening earns attention. Hold rate takes the people it earned and asks what happened next. The hook's effect is divided out, so what remains describes the body of the ad.

This is why the two metrics belong together. Either alone leaves the important question open; the pair answers it.

High hook, high hold. The opening earned attention and the video repaid it. This is the creative to scale, and the thing to watch for is fatigue rather than performance.

High hook, low hold. The opening stopped people and the rest lost them. The two usual causes are a hook that promised something the body did not deliver, and a video longer than the idea inside it.

Low hook, high hold. The most actionable pattern in video creative. The body works and almost nobody is getting to it. A new opening against the same footage is a cheap edit with a large effect, and it is the highest-return change available in most video accounts.

Low hook, low hold. Usually not a creative problem, particularly if it holds across several different creatives. Check the placement mix and the audience before commissioning anything new.

Hold rate compresses a distribution into one number, and the distribution is where the instructions are.

A sharp drop at a specific moment names something: a scene change, a price, a claim, a shift in tone, the point where the ad becomes visibly an ad. That moment is locatable and editable.

A gentle, even decay usually means the video is simply longer than the idea in it. A shorter cut will hold better and frequently converts better too.

A plateau followed by a cliff is the shape of an ad that works until it asks for something, and the cliff is where the ask lands.

An early collapse that then flattens means the ad lost the wrong audience quickly and kept the right one, which can be a perfectly good outcome and looks poor in the ratio.

Reading the curve takes a minute and produces an edit. Reading the ratio produces a decision to make another video.

The figure depends on both thresholds, so comparisons only work when both ends are held constant.

Comparing a three-to-fifteen hold rate against a three-to-completion one is comparing different quantities. So is comparing a fifteen-second video's hold rate against a sixty-second video's, since holding somebody for fifteen seconds of a fifteen-second video and fifteen seconds of a minute-long one are different achievements.

Placement matters as much here as it does for hook rate. Autoplay feed inventory produces a different retention shape from placements where somebody pressed play, so a blended figure moves when the delivery mix moves. Splitting by placement before concluding is the check that prevents most false results.

It does not measure comprehension. Somebody can watch to the end and take nothing from it.

It does not measure persuasion. Attention is necessary and not sufficient, and the link between watching and acting is loose enough that neither rate can stand in for an outcome.

It does not measure sound. A silent viewer is counted like any other, and a video whose message lives in the audio can hold attention and communicate nothing.

And it is not a target. Optimising toward it directly produces videos engineered to retain rather than to sell, which is a real failure mode in accounts that have discovered the metric recently.

In the creative review, beside hook rate and the retention curve, answering the question the outcome metrics cannot: given that the ad is underperforming, is it failing at the opening or after it.

Those are different problems. The first is fixed with a new first second; the second with a cut, a restructure or a shorter video. An account that reads only cost per result will attempt both at random, and will frequently rewrite an offer that nobody stayed long enough to hear.

The operational argument for watching this pair is speed.

Conversions accumulate slowly. A creative test judged on cost per conversion needs weeks in most accounts and may never reach significance in a small one. Retention figures accumulate within days, because every impression contributes to them.

That makes hook rate and hold rate the practical signals for early creative decisions — not because they are what the business cares about, but because they arrive in time to act on. The discipline is to use them to shortlist rather than to decide: retire the clearly weak quickly on retention, and let the survivors run long enough to be judged on outcomes.

An account that judges creative only on conversions tests very little, because each test costs weeks. One that judges only on retention tests a great deal and optimises for watching rather than buying. The pair of habits together is what works.

Hold rate flatters short videos and punishes long ones, mechanically.

Holding somebody for fifteen seconds of a fifteen-second video is completion; holding them for fifteen seconds of a ninety-second video is an eighth of the way in. The same ratio describes two very different achievements.

Where an account runs a mix of lengths, comparing hold rates across them is comparing the lengths rather than the edits. Either compare within a length band, or use the proportion watched rather than a fixed threshold, which normalises the comparison at the cost of being less directly available in the reporting.

• Find where the retention curve falls and cut there. A curve that collapses at a specific moment names a scene, a claim or a tonal shift that is losing people.
• Shorten before restructuring. A gently decaying curve usually means the video is longer than the idea in it, and a cut-down edit is cheaper than a rewrite.
• Make sure the body delivers what the hook promised. A high-stopping opening followed by an unrelated middle produces exactly this failure.
• Keep the denominator fixed when comparing edits, or the comparison measures the opening rather than the body.
• Read it beside hook rate rather than alone, because the pair separates an attention problem from a retention one.