Glossary2026-08-08Reference
Hold rate is ThruPlays over 3-second plays: what the hook handed over and what the body kept. Formula, worked example, calculator, and how Admetra caps a body beat.
Hold rate is the share of the people the hook delivered who then stayed: ThruPlays divided by 3-second video plays. A ThruPlay is a play of at least 15 seconds, or of at least 97% of the ad if it runs shorter than 15 seconds. On a 12-second ad hold rate is a completion rate, and on a 40-second ad it is a 15-second rate. Compare it across ads of similar length, or you are comparing two different measurements. Where hook rate grades the opening, hold rate grades everything after it: the body, the pacing, the proof. The two fail in opposite directions and want opposite fixes: a high hook rate with a low hold rate means the opening wrote a check the body did not cash, and re-shooting the hook makes that ad worse, not better.
Hold rate = ThruPlays ÷ 3-second video plays × 100. Both terms are Meta's: TikTok's video play metrics list 2-second and 6-second video views and no 3-second play, and the nearest thing to a ThruPlay - 15-second focused views - is available only to advertisers on focused view products. A TikTok hold rate is built from different thresholds and is not comparable to this one. Some teams divide by impressions instead, which produces a much smaller number that blends hook and body into one score and can no longer tell you which of the two is broken. If you keep only one definition, keep this one. The denominator has to be the audience the hook actually delivered.
An ad records 12,600 three-second plays and 3,150 ThruPlays: a 25% hold rate. A second ad in the same campaign records 5,000 three-second plays and 2,500 ThruPlays: 50%, on less than half the traffic. The first ad has a stronger opening and a leaking middle; the second has the opposite problem. Spending the next production day on the wrong one of those is the most common way a creative test wastes a week.
Hook rate is 3-second video plays divided by impressions. Put your own two numbers in.
7.86%hook rate
Long ads lose people in flat stretches: a single undifferentiated middle where nothing escalates. Admetra's analysis is built to make those stretches visible: a body beat is capped so that a long, repetitive section cannot be recorded as one block, longer videos are required to produce more beats rather than broader ones, and on a long ad a stretch of timeline with no beat assigned to it is treated as a gap in the analysis, not as a part of the ad.