Hold rate: How to calculate it on Meta, TikTok, and YouTube

Hold rate shows whether viewers keep watching after the hook. The formula on Meta, TikTok, and YouTube, how ad length changes it, and how to find the drop.

Hands hold a phone playing a starry desert video ad with a coral tent while laundry spins in a washing machine in a daylight laundromat.
On this page
  1. The hold rate formula
  2. Hold rate on TikTok and YouTube
  3. How ad length changes hold rate
  4. Finding where viewers drop
  5. Reading hold rate with hook rate
  6. How to calculate hold rate in a report
  7. Locating the weak scene with Smart Insight and Video Agent
  8. Frequently asked questions

Hold rate, or video hold rate, is the share of viewers who keep watching a video ad after its opening. On Meta, hold rate is ThruPlays divided by 3-second video plays, times 100: of the viewers who watched the first 3 seconds, it counts how many stayed to the 15-second mark, or to nearly the end of a shorter video. Hold rate is built from each platform's own view counts, so the number means little until the report names the two metrics behind it.

Next to hook rate, which measures the opening, hold rate shows whether an ad loses its viewers in the first scene or in the scenes after it.

The hold rate formula

On Meta, the formula is:

Hold rate = ThruPlays ÷ 3-second video plays × 100

An ad with 12,000 three-second video plays and 3,600 ThruPlays has a 30% hold rate. Both inputs are Meta's own video ad metrics:

  • Three-second video plays: the denominator, "the number of times your video plays for at least 3 seconds, or for nearly its total length if it's shorter than 3 seconds," counted separately for each impression, with replays excluded.
  • ThruPlays: the numerator, "the number of times your video was played to completion, or for at least 15 seconds." Per Meta's ThruPlay page, a video shorter than that earns one "after at least 97% of the video was played." Meta also offers ThruPlay at 6 seconds.

Dividing ThruPlays by impressions instead gives a different number, one that folds the hook into the hold: with a 28% hook rate, a 25% hold rate on 3-second plays equals 7% of impressions. State the denominator next to the name in every shared report.

Keep both inputs on the same counting method. In Meta's calculation notes, 3-second video plays and ThruPlays sum "the unique seconds that a video plays." Video average play time is counted another way: it includes replays, and Meta's average play time page notes the reported average "could be longer than the length of the video."

Hold rate on TikTok and YouTube

TikTok's video play metrics, last updated in November 2025, list no metric named ThruPlay. They give these inputs:

  • Two-second video views: plays of "at least 2 seconds in an impression session," with replays excluded.
  • Six-second video views: plays of at least 6 seconds, or the full video if it is shorter, plus any view that "received at least 1 engagement within the first 6 seconds."
  • Video views at 25%, 50%, 75%, and 100%: plays that reached each share of the video's length, counted separately for each impression, with replays excluded.

A TikTok hold rate built like Meta's divides video views at 50% by 2-second video views. On a 30-second ad, the 50% mark falls at 15 seconds, the same point a Meta ThruPlay measures. Six-second views over 2-second views give a quicker read of the shots that follow the hook, but any engagement in the first 6 seconds also counts as a 6-second view, so that ratio mixes watching with engagement.

YouTube reports its quartiles as rates already, so no division is needed. Google Ads' "video played to" metrics "count the percentage of people who watched 25%, 50%, 75%, or 100% of the video out of the viewers who initiated the player." Video played to 50% is the YouTube counterpart of hold rate, with player starts rather than early views as its denominator.

Skippable ads have a known drop point. Google adds that "it's common to find a steep drop off of users after 5 seconds for in-stream ads when the skip button appears for users." On a 20-second skippable ad, the 25% mark falls at that skip point.

A Meta hold rate and a TikTok hold rate of the same value still describe different behavior. Meta's early threshold is 3 seconds and TikTok's is 2, so a fair comparison stays inside one platform and one placement.

How ad length changes hold rate

ThruPlay is set at a fixed time, not a share of the ad, so the same metric reads differently by length:

  • Up to 15 seconds: a ThruPlay needs the whole video, or 97% of it when the video is shorter than 15 seconds, so hold rate is a completion rate.
  • Thirty seconds: a ThruPlay needs 15 unique seconds played, half the ad for a viewer who watches from the start, so hold rate measures early retention.
  • Sixty seconds: a ThruPlay needs the same 15 seconds, a quarter of the ad, and the scenes after it barely move the number.

Group ads into length bands before comparing them, for example 15 seconds or shorter, 16 to 30 seconds, and over 30 seconds, and set a separate median for each band. With the 6-second ThruPlay, only ads of 6 seconds or shorter fall into the completion band, and longer ads up to 15 seconds move into early retention.

Looping changes a different number. Instagram videos "always loop as long as visible," according to Meta's video length and looping page, and video average play time counts those replays while ThruPlays do not.

On YouTube, length changes the view itself. A skippable in-stream view counts at "30 seconds or until the end of the video, whichever comes first," or on a click. That favors a shorter ad: Google's campaign tips state that "Shorter ads have higher TrueView view rates." Compare a shorter edit with the original at the same second, using the closest quartile.

Finding where viewers drop

The quartile counts locate the drop, and the shot list names it. Four steps turn one into the other:

  1. List the shots with timestamps: the opening, each scene, and the end card.
  2. Convert each quartile to seconds: on a 20-second ad, 25% falls at 5 seconds, 50% at 10, and 75% at 15.
  3. Compute retention between points: each count divided by the one before it.
  4. Mark the largest drop: the shot that spans it is the one to revise first.

On Meta, use the quartile counts for location and the ThruPlay ratio for the headline number. Meta's video plays at 25% "may also include plays from people who skipped to the 25% point," while ThruPlays count unique seconds watched.

Example: Reading the drop in a 20-second app install ad

Illustrative counts for a 20-second app install ad on Instagram Reels, to replace with the shot times and counts from an export:

  • Shot list: gameplay hook from 0 to 3 seconds, feature demo from 3 to 8, a user's reaction from 8 to 13, the offer from 13 to 17, and the end card from 17 to 20.
  • Counts: 12,000 three-second video plays, 9,600 plays at 25% (5 seconds), 5,800 at 50% (10 seconds), 4,900 at 75% (15 seconds), 3,700 at 100%, and 4,700 ThruPlays.
  • Hold rate: 4,700 ÷ 12,000 = 39.2%.
  • Retention between points: 80% from 3 to 5 seconds, 60% from 5 to 10 seconds, 84% from 10 to 15 seconds, and 76% from 15 seconds to the end.
  • Read: the largest drop falls between 5 and 10 seconds, across the end of the feature demo and the first seconds of the reaction shot.
  • Next test: a re-edit of the feature demo that reaches the result sooner, with every other shot unchanged, judged on the 50% count and on hold rate.

Reading hold rate with hook rate

A low hook with a high hold points to the first scene, and a high hook with a low hold points to the scenes after it. All four combinations and the fix for each are in Starti's guide to hook rate. Hold rate also feeds the read in creative analytics, where it sits next to cost per result.

How to calculate hold rate in a report

Build hold rate from summed counts, not from an average of per-ad rates. In Looker Studio, a calculated field can divide one aggregated metric by another, following Google's own example, SUM(Profit) / SUM(Revenue):

  • Formula: SUM(ThruPlays) / SUM(3-second video plays), formatted as a percent. Field names depend on the data connector.
  • Name: the formula in words, such as "Hold rate (15s ThruPlays ÷ 3s plays)," so anyone reading the report knows the inputs.
  • Meta Ads Manager: the same ratio as a custom metric, built from Customize columns, then Create custom metric.
  • TikTok: SUM(video views at 50%) / SUM(2-second video views), in its own field.
  • YouTube: the video played to 50% column is already a rate, so report it at the level being compared instead of averaging it across rows.

Averaging per-ad rates gives a small ad the same weight as a large one. Two ads show the gap: ad A has 1,000 three-second plays and 500 ThruPlays, a 50% hold rate, and ad B has 20,000 and 4,000, a 20% hold rate. The average of the two rates is 35%, while the summed counts give 4,500 ÷ 21,000, or 21.4%. Looker Studio sets a limit of its own: "Re-aggregating metrics is not supported," according to its calculated field troubleshooting page, which says this includes most metrics found in Google Analytics and Google Ads.

Locating the weak scene with Smart Insight and Video Agent

At Starti, campaign findings decide which hooks, scenes, and messages stay in the next brief. Smart Insight compares nine creative attributes against CTR, completion, and other outcomes across Meta, TikTok, and Google Ads. Each finding shows supporting creatives, counterexamples, and sample size, with the timestamped scenes in view.

When the drop points to one scene, Video Agent in Starti AI Studio revises that scene and keeps the rest of the cut exactly as approved, and the next test changes just the scene that lost viewers.

Book a demo to see Smart Insight and Video Agent.

Frequently asked questions

What is hold rate?

Hold rate is the share of viewers who keep watching a video ad after its opening. On Meta, it is ThruPlays divided by 3-second video plays, times 100.

How do you calculate hold rate?

Divide a later view count by an early one and multiply by 100: ThruPlays over 3-second video plays on Meta, and video views at 50% over 2-second video views on TikTok. YouTube's video played to 50% column already reports it as a rate.

What is a good hold rate?

The useful comparison is the account's own median for ads in the same length band, placement, and objective over the last 30 days. Flag ads in the bottom quarter of their band for a re-edit of the scene where the largest drop falls.

Is hold rate the same as completion rate?

Only for ads of 15 seconds or shorter, where a ThruPlay needs the whole video, or 97% of it for shorter videos. On a 30-second ad, it measures the first 15 seconds watched.

Starti AI Studio

Discuss your next campaign with Starti.

Book a Demo
Hands pack ski mittens and goggles into an open backpack while a television behind a low partition plays a ski resort ad featuring a distant skier in a coral jacket.

TV attribution: How it works for linear and streaming TV

TV attribution explained: how linear and streaming TV ads get credit for visits and sales, what OTT attribution covers, and how TV attribution windows work.

A folded gray stroller stands in a sunlit entryway while a television behind a low plaster partition shows a stroller with a coral canopy on a park path.

Programmatic TV advertising: How it works and how to buy it

Programmatic TV advertising explained: how one streaming TV impression is bought, how linear TV differs, programmatic vs direct buying, and how to start.

A hand selects a sailing shot from a row of hanging footage stills, in navy halftone with a coral rectangle behind the chosen print.

Video asset management: How ad teams find and reuse footage

Video asset management for ad teams: why video is harder than images, indexing by shot, reusing footage, the workflow to publishing, and what software needs.