What is ThruPlay?

A twelve-second video delivered 300,000 times records 84,000 two-second views and 9,000 ThruPlays. Because the video is under fifteen seconds, a ThruPlay means it was watched to the end: 9,000 ÷ 84,000 = 10.7% of the people who started got to the end, and 3% of the impressions produced one. The first ratio is a fact about the edit; the second is a fact about the edit and the inventory together.

ThruPlay counts a video play that reached the end, or that lasted at least fifteen seconds where the video runs longer than fifteen seconds.

The definition is two rules in one, and which rule applies depends on the length of the video. For anything under fifteen seconds, a ThruPlay is a completion. For anything longer, it is a fifteen-second watch with no statement about what happened afterwards.

That asymmetry is worth holding onto, because it means the metric is not measuring the same thing across a mixed set of creatives. A six-second ad and a sixty-second ad both report ThruPlays and the underlying events are quite different.

The metric replaced a shorter default, and the reason is instructive about how optimisation metrics behave in general.

In-feed video autoplays silently. A person scrolling at ordinary speed crosses a two- or three-second threshold constantly without any decision to watch. When that threshold is what a campaign is optimised toward, the bidder learns to find inventory and audiences that generate it cheaply — which means fast scrolling past autoplaying video.

The campaign report looks excellent. The cost per view is tiny, the view count is enormous, and nobody has watched anything.

Raising the threshold changes what the bidder is looking for. Fifteen seconds of attention cannot be produced by scrolling, so the model has to find people who stop. The events cost several times more and they are events.

This is the general lesson: an automated bidder optimises the metric it is given with no regard for what the metric was meant to represent. Where the metric is easy to produce accidentally, the bidder will produce it accidentally, at scale.

It proves attention was held. Fifteen seconds is long enough that a person was watching rather than passing.

It does not prove the message landed. Attention is necessary and not sufficient, and a video that held fifteen seconds while saying nothing useful has bought nothing.

It does not prove the whole video was seen, on anything over fifteen seconds. The tail beyond the threshold is unmeasured by this metric and needs the completion rate instead.

It does not prove sound was on. Silent watching is still watching, and a video whose message lives in the audio can accumulate ThruPlays and communicate nothing.

The absolute count is a function of budget and inventory price. The ratio against the shortest available view threshold is a function of the video, and it is the number that carries instruction.

If eighty-four thousand people crossed the two-second mark and nine thousand reached the end, roughly one in nine of the people who started finished. That figure responds to editing and almost nothing else, which makes it the cleanest creative test available on video.

Tracking it across a set of creatives ranks them by holding power. Tracking it across versions of the same creative — a different opening, a shorter cut, captions added, the point moved earlier — measures the edit directly.

The decision follows the creative rather than a default.

Choose it when the message needs time: a demonstration, a before and after, an explanation, a story with a turn in it. In those cases a two-second view is worthless and paying several times more for a real one is the correct trade.

Do not choose it when the entire message lands immediately: a price, an offer, a single striking image with three words on it. Paying for fifteen seconds of attention on a creative that finishes its work in two buys nothing and restricts delivery.

Where it is chosen, expect the reported cost per event to be several times the cost of a short-threshold view, and expect volume to be much lower. Neither is a problem to be optimised away; both are what the metric was designed to produce.

A few consequences follow directly.

Keep videos short enough that a ThruPlay means completion. Under fifteen seconds, the metric is unambiguous, and the discipline of fitting the idea into that window usually improves the ad anyway.

Front-load the point. The threshold measures whether people stayed; what keeps them is knowing within the first moments that there is a reason to.

Assume silence. Captions carry the message; audio decorates it.

Do not pad to reach fifteen seconds. A twelve-second video that finishes is better on this metric than a sixteen-second one that does not, because the first records completions and the second records abandonment.

ThruPlay sits between the loose view thresholds and full completion rate, and the three together describe the shape of attention a video earned.

The short threshold says how many people were served and did not immediately scroll. ThruPlay says how many were still there when the message had time to land. Completion says how many saw the whole thing.

A campaign strong on the first and weak on the second has a hook that works and content that does not. Strong on the second and weak on the third has a middle that sags. Weak on all three is usually an inventory problem rather than a creative one, and the fix is a placement change before an edit.

The people who produced a ThruPlay are a usable audience, and this is frequently worth more than the metric itself.

Somebody who stayed fifteen seconds with an ad has self-selected in a way no impression-based audience can match. Retargeting pools built from that threshold are small, warm and cheap to build, and in many accounts they perform closer to site-visitor audiences than to broad prospecting.

The practical sequence is to run video against a broad audience with this metric as the objective, then retarget the people it qualified with a direct-response campaign. That structure turns upper-funnel video from a spend with a distant payoff into one that produces an asset within a week, which is the strongest argument for buying video in an account that measures itself on sales.

The threshold used to build the pool matters. A pool built from two-second views contains everyone who scrolled past and is barely warmer than the campaign's reach. A pool built from this threshold contains only people who stayed.

• Use it as the optimisation event when the message needs more than a glance, and accept that the delivered cost will be higher than a short-threshold objective produces.
• Keep videos under fifteen seconds where the point allows, so that a completion and the threshold are the same event and the number is unambiguous.
• Read it as a ratio against the shortest view threshold. That ratio is retention and it is what the creative controls.
• Do not compare it with another platform's completion metric. The fifteen-second rule and the completion rule are different definitions living under one name.
• Check it after every edit rather than on a schedule. The number responds to cuts faster than to anything a media setting can do.