Meta Ads video: 10.5M plays, 4.8% reached the end, 5.9% counted as a ThruPlay
Across 23 Australian Meta accounts and $783,518 (AUD) of pooled lifetime spend, video ads recorded 10,523,670 plays. 504,102 of those plays reached the end of the video, which is 4.8%. Meta's ThruPlay metric, the one most video campaigns optimise for and report on, counted 618,373, or 5.9%.
The gap between the first number and the last is the subject of this post. A "play" in Meta's video_play_actions is any start, including the autoplay that begins the moment a video scrolls into the feed. Nobody chose to watch. A ThruPlay is a play that either reached the end or reached 15 seconds, whichever comes first. So the 10.5 million is not an audience, and the 5.9% is not a completion rate, and the number that comes closest to "people who saw the whole ad" is the smallest one on the page.
We pulled the full watch funnel: plays, then the share reaching 25%, 50%, 75% and 100% of the video's length, and ThruPlay alongside. Eleven accounts returned video plays in meaningful numbers; the pooled figures cover the whole set.
Stage | Plays | Share of plays |
|---|---|---|
Play (any start, including autoplay) | 10,523,670 | 100% |
Reached 25% | 2,372,486 | 22.5% |
Reached 50% | 1,228,295 | 11.7% |
Reached 75% | 760,759 | 7.2% |
Reached 100% | 504,102 | 4.8% |
ThruPlay (15 seconds or complete) | 618,373 | 5.9% |
The steepest loss is at the first step. 77.5% of plays never reach a quarter of the video, which for most ads means the viewer scrolled past inside the first few seconds. After that the funnel roughly halves at each quarter: 22.5% to 11.7% to 7.2% to 4.8%. The ThruPlay figure sits above the completion figure because it includes 15-second watches of longer videos that did not run to the end. Subtracting one from the other, 114,271 ThruPlays reached 15 seconds without reaching the end.
Why this happens
Meta's video metrics are defined generously at the top and honestly at the bottom, and the interface encourages you to read the top.
Metric | API field | Counts when | What it is not |
|---|---|---|---|
Video plays |
| The video starts, by any means including autoplay | A person choosing to watch |
3-second plays |
| 3 seconds or more (not pulled for this pack) | A view in any other platform's sense |
ThruPlay |
| Played to completion, or to 15 seconds if the video is longer | A completion |
25% / 50% / 75% |
| Playback reached that fraction of the video length | Comparable across videos of different length |
100% |
| Playback reached the end | Necessarily attention; the tab may be open with the sound off |
Autoplay in feed is the reason the play count is so large. Meta starts a video when it enters the viewport, on mobile and desktop, without a tap. On the 94.5% of this set's spend that went to phones, that means a play is registered for nearly every impression where the video was on screen long enough to begin. A play is closer to a viewable impression than to a view.
ThruPlay was introduced as the fix for that. It was meant to be a metric that reflected a real watch, and for a video under 15 seconds it does: a ThruPlay is a completion. For a 60-second video it is a quarter of the way through. The pull does not contain video lengths, so we cannot say how many of the 618,373 ThruPlays are completions of short videos and how many are 15-second slices of long ones. What we can say is that in every one of the eleven accounts, the ThruPlay rate exceeds the completion rate, which means every account is running at least some video longer than 15 seconds.
The percentile metrics are the useful ones and the least used. They are relative to each video's own length, so a 50% watch of a six-second video and a 50% watch of a two-minute video are both counted as 50%. That makes them hard to compare across videos and easy to compare across audiences for the same video. They are also the thresholds Meta lets you build engagement audiences on, which is where they matter most.
What it does to decisions
The first decision is what to optimise for. A video campaign set to the ThruPlay optimisation goal asks Meta to find people likely to watch 15 seconds or to the end. Meta is good at that. It will find people who leave videos running, and it will report a ThruPlay rate that looks healthy. Whether any of those people were in the market is a separate question that the metric does not address. In lead generation accounts, which most of this set are, a video campaign optimised for ThruPlay produces ThruPlays, and the lead comes from a different campaign or not at all.
The second is what to put in the video. If the call to action is at the end, 4.8% of plays saw it. If it is at the halfway mark, 11.7%. If it is in the first three seconds, most of the 10.5 million had a chance at it. The completion funnel is a map of where the message can go.
The third is what audience to build. Meta's engagement audiences let you retarget people who watched 3 seconds, 10 seconds, 15 seconds (ThruPlay), 25%, 50%, 75% or 95% of a video. The table below shows what each of those pools looked like in this set, using the percentile thresholds we pulled.
Retargeting threshold | Pooled count | Share of plays | What it selects for |
|---|---|---|---|
Any play | 10,523,670 | 100% | Being on the feed when the ad loaded |
25% watched | 2,372,486 | 22.5% | Did not scroll past immediately |
50% watched | 1,228,295 | 11.7% | Some interest, or a short video |
75% watched | 760,759 | 7.2% | Sustained attention |
100% watched | 504,102 | 4.8% | Saw the whole thing, including the end card |
An audience built on any play, or on 3-second views, is not much narrower than the reach audience it came from. An audience built on 75% is a twentieth the size and is made of people who chose to keep watching. The mistake we see most often is the first one: a "video viewers" retargeting audience that is effectively "people who were shown the ad".
The same definitional problem exists on the Google side, with different thresholds. YouTube counts a view on a skippable in-stream ad at 30 seconds, at completion for shorter ads, or on interaction, so a YouTube view rate and a Meta ThruPlay rate are measuring different lengths of attention. We covered the Google version in YouTube ads for lead generation. Neither platform's headline video metric is comparable to the other's, and neither is a completion rate.
Where it varies
Here is every account with video plays, sorted by completion rate.
Acct | Band | Plays | 25% | 50% | 75% | 100% | ThruPlay |
|---|---|---|---|---|---|---|---|
B | Manufacturing & equipment | 82,949 | 49.1% | 30.6% | 21.2% | 14.7% | 18.6% |
R | Manufacturing & equipment | 42,705 | 39.2% | 23.4% | 16.5% | 11.9% | 12.4% |
G | Home improvement & trades | 1,113,079 | 30.0% | 19.5% | 14.4% | 10.5% | 13.6% |
H | Health & care services | 288,854 | 27.3% | 17.4% | 12.2% | 10.0% | 13.1% |
F | Manufacturing & equipment | 2,594,345 | 32.7% | 16.8% | 10.2% | 6.1% | 6.3% |
M | Home improvement & trades | 40,491 | 19.5% | 11.5% | 8.4% | 5.7% | 7.0% |
A | Consumer goods & food | 3,808,354 | 23.0% | 10.4% | 6.3% | 4.2% | 5.0% |
S | Consumer goods & food | 237,079 | 19.7% | 9.0% | 3.2% | 1.9% | 4.9% |
P | Home improvement & trades | 949,204 | 9.6% | 3.0% | 1.6% | 1.0% | 1.4% |
C | Manufacturing & equipment | 186,533 | 8.4% | 2.8% | 1.3% | 0.8% | 2.0% |
Q | Health & care services | 1,178,056 | 1.4% | 2.9% | 0.6% | 0.3% | 2.0% |
The completion rate runs from 0.3% in account Q, a health services advertiser with 1.2 million plays, to 14.7% in account B, a manufacturer with 82,949. That is the same metric, on the same platform, in accounts pulled on the same day, and the range is wide enough that no pooled figure describes any one of them.
Volume explains the pooled figure. Accounts A and F between them hold 6,402,699 of the 10.5 million plays, and their completion rates are 4.2% and 6.1%. The pooled 4.8% is mostly those two accounts. The high performers are small: B and R together have 125,654 plays against account A's 3,808,354, and their 14.7% and 11.9% move the pooled figure very little. The eleven rows sum to 10,521,649 plays; the remaining 2,021 in the pooled total sit in accounts with too few plays to show.
Band is a weak guide. Manufacturing holds the two best accounts, B and R, and one of the three worst, C at 0.8%. Home improvement spans G at 10.5% and P at 1.0%. Health spans H at 10.0% and Q at 0.3%. The pull does not tell us video length, placement mix or objective per account, and any of the three could produce a spread this wide on its own. A six-second cutdown placed in Stories will complete at rates a two-minute explainer in Reels never will, and neither number says which was the better ad.
One row needs a flag. Account Q reports 1.4% of plays reaching 25% and 2.9% reaching 50%. In a single video's funnel the second cannot exceed the first. We have reported the figures as pulled; the likely cause is that the account's percentile metrics are pooled across videos of very different lengths and delivery periods, and Meta's percentile counters are not guaranteed to be consistent with each other at account level. Treat account Q's funnel as unreliable above the 100% row, which at 0.3% is the lowest in the set either way.
What to check in your account
In Ads Manager, on the Ads tab, choose the Video engagement column preset. It adds plays, 3-second plays, ThruPlays, and the 25% through 100% watch columns.
Filter to video ads and set the range to lifetime. Sort by video plays, descending.
For your top five ads by plays, divide 100% watches by plays. That is the completion rate. Compare it to the ThruPlay rate in the next column; the gap is the share of 15-second watches on videos longer than 15 seconds.
Note the length of each of those videos. A ThruPlay on a 12-second video and a ThruPlay on a 90-second video are not the same event.
Open Audiences and find every video engagement audience. Note the threshold each one uses. Anything on 3 seconds or any play is close to a reach audience.
Check the optimisation goal on every video campaign. If it is ThruPlay in a lead generation account, ask what the campaign is for.
Look at the 25% column. If most plays do not reach it, the message has to be in the first few seconds or it is not being delivered.
What you see | What it means | What to do |
|---|---|---|
Plays in the millions, completions in the tens of thousands | Autoplay is doing the counting | Report on 75% or 100%, not plays |
ThruPlay rate well above completion rate | Long videos, 15-second watches counted | Note video length beside every ThruPlay figure |
ThruPlay rate equal to completion rate | Videos under 15 seconds | ThruPlay is a completion here and can be read as one |
25% watch rate under 10% | The first few seconds are losing nearly everyone | Move the message or the hook earlier |
Retargeting audience built on any play | Effectively a reach audience | Rebuild on 50% or 75% and accept the smaller pool |
Percentile columns not monotonic | Pooled across videos, counters inconsistent | Read per ad, not per account |
We check the video funnel on every account that runs video, because the play count is the number that gets quoted and the completion count is the number that matters. Our Facebook Ads management reports the whole funnel, with video length beside it.
The principle
A play is a video starting, not a person watching. A ThruPlay is fifteen seconds, or the end, whichever comes first, and on a long video that is not much. The only column that means what it sounds like is the last one, and it is one in twenty. Report the funnel, not the top of it, put the message where the funnel says the audience still is, and build audiences on the people who stayed rather than the people who were shown.
