What is Thumbstop rate?

A video is served 350,000 times to 140,000 people and records 38,500 three-second plays. Over impressions the thumbstop rate is 38,500 ÷ 350,000 = 11%. Over reach it is 38,500 ÷ 140,000 = 27.5%. Both are called thumbstop rate, and the gap between them is exactly the campaign's frequency of 350,000 ÷ 140,000 = 2.5.

Thumbstop rate is the share of impressions where somebody stopped scrolling long enough to register the ad, computed from a short-threshold video view count.

Thumbstop rate = three-second video plays ÷ impressions

That is the same arithmetic as hook rate in most accounts that use both terms. Where a distinction is claimed, it is one of emphasis rather than of computation: thumbstop implies a scrolled feed and the physical act of a thumb pausing, while hook implies the opening of the creative and works on surfaces where nobody is scrolling at all.

The practical advice is short. If an account uses both terms, write down which is which. The failure mode is not disagreement about the numbers; it is the same number appearing twice in a report under two headings and somebody concluding the figures disagree.

The genuine ambiguity in this metric is not the name but the denominator, and the two versions differ by a large factor.

Over impressions, the figure describes deliveries: of the times the ad was served, how many produced a stop.

Over reach, it describes people: of the people served, how many stopped at least once.

The ratio between them is the campaign's frequency. At a frequency of 2.5, the reach version is two and a half times the impression version from exactly the same campaign.

The case for reach is that a person served five times had one chance to be interested rather than five. The case for impressions is stronger for the metric's main use: it is stable against frequency, so it changes when the creative changes rather than when delivery does. Since this figure exists to compare creatives, the version that does not move for unrelated reasons is the better instrument.

Either is defensible. Neither is defensible unstated.

The mechanisms are physical before they are persuasive.

Motion. A moving first frame interrupts a scroll in a way a still does not. Opening on a static shot gives up the main thing the format offers.

A face. Attention goes to faces faster than to products, which is most of why creator-style openings outperform product-first ones on feed surfaces.

Contrast against the feed. An ad that looks like the surrounding content gets less attention than one that breaks the visual pattern — though breaking it too obviously reads as an advertisement, which is its own reason to scroll.

Recognition. The most durable mechanism: the viewer sees their own situation and stops because it is about them.

Text that can be read in a glance. A few large words legible at speed do work that a sentence cannot.

A meaningful share of any thumbstop figure is not a stop at all.

In-feed video plays automatically and silently. Somebody scrolling at ordinary speed crosses a two- or three-second threshold without deciding anything, and the platform counts it.

This inflates the metric on exactly the inventory where it is most often measured, and it is why a very high figure on cheap feed placements should be read sceptically rather than celebrated.

The correction is to read it beside a retention measure. A stop that does not persist was not attention worth buying, and the pair separates a real stop from a slow scroll far better than either number alone.

The figure varies more by surface than by opening.

Full-screen vertical placements produce different rates from in-feed squares, which produce different rates from in-stream placements where the person chose the content. Autoplay behaviour differs by surface. Player size and position differ.

So a blended thumbstop rate across placements is a weighted average of several behaviours, weighted by a delivery algorithm that changes the mix continuously.

Any creative comparison made on the blended figure is suspect, and splitting by placement before concluding is the single check that makes the metric trustworthy. A creative that appears to have won frequently received more of a favourable surface.

Test first frames, not whole ads. The thumbnail moment is most of what the metric measures, and it is the cheapest thing in a video to change.

Hold the placement constant when comparing, or split the report by it.

Write the definition into the report template, threshold and denominator both, so that the number means the same thing next quarter.

Never optimise toward it. Buying cheap autoplay impressions raises it without buying attention, and a campaign pointed at this figure will find exactly that.

Read it as one of a pair. The stop is only interesting if something follows it, and the account's real question is always whether the ad earned the second that came after the one it bought.

The idea transfers to static creative, where no view threshold exists to compute it from.

A static ad has the same job in the same moment: the first glance decides whether anything else is read. What it lacks is a metric, because there is no three-second view count for an image.

The available proxies are weaker. Click-through rate mixes the stop with the offer and the call to action. Engagement rate mixes it with everything that provokes a reaction. Neither isolates the moment the way a short-threshold view does.

The practical consequence is that static creative has to be tested on outcomes rather than on attention, which is slower and needs more volume. It is also an argument for producing video versions of static ideas where the budget allows, since the additional diagnostic is worth something beyond the format itself.

Treated correctly, this is a fast, cheap signal about one specific thing: whether an opening earns a moment.

Treated as a performance measure it is actively misleading, because the cheapest way to raise it is to buy the inventory that produces accidental stops.

The account's position should be that it never appears in a performance summary, always appears in a creative review, and always carries its definition. Those three rules cover essentially every way the metric goes wrong.

• Write the denominator beside the number. The impression and reach versions of this figure differ by the campaign's frequency and both carry the same name.
• Rebuild the first frame rather than the first sentence. The thumbnail moment is what a scroll sees, and a still that works is most of the job.
• Open on motion, because motion is the mechanism that interrupts a scroll and a static first second gives it up.
• Compare creatives within one placement. Autoplay surfaces and chosen-content surfaces produce different figures from identical openings.
• Read it against hold rate, because a scroll that stops and immediately resumes was not attention worth buying.