Insights - Brand

Short-form video: stop buying views, buy entry points

Short-form video is the cheapest machine ever built for attaching a brand to a buying occasion, and it is sold by the one unit that cannot tell you whether it did. The platforms know. They do not use that unit when the money is theirs.

A category entry point is the cue a buyer thinks from when they enter a category: the situation, the occasion, the moment of need, not the product. The Ehrenberg-Bass Institute, whose researchers named the idea, puts it plainly - these are “moments when someone mentally enters a product or service category”, and linking a brand to as many relevant cues as possible is what increases mental availability. Short-form video is the best medium ever built for that job and the worst-measured. The verdict: short-form belongs in the brand budget, bought and judged on coverage of entry points, and the reason almost nobody runs it that way is that the unit on the media plan ends the moment the video starts playing.

Why short-form is structurally an entry point medium

Look at the grammar of the two formats. A thirty-second television commercial opens on a product or a brand world, because it was bought in a slot the viewer agreed to sit through. A feed video opens on a situation, because it has to earn the next second: a kitchen at eleven at night, a gym bag by the door, a kid's party going wrong, Sunday evening and nothing in the fridge. That is not a creative fashion. It is what the format rewards, and it happens to be the exact raw material of an entry point.

Which hands you an unusual opportunity and an easy trap. The opportunity is that you can cover a dozen separate occasions in a quarter for the cost of one television execution, and occasions are what grow a brand, because each one is a separate door into the category. The trap is that the format rewards the situation and the brief usually specifies the product, so the video that performs best in the feed is often the one that covers no entry point at all. It covers the category.

Hold that thought, because it is where the money leaks, and we will come back to it with an arithmetic cost attached.

The unit you buy ends at the start of play

Read the platforms on their own metrics. TikTok's help centre defines video views as the “number of times your video started to play”. Plays are counted separately for each impression and replays are excluded, which is honest bookkeeping of the thing it is counting. It is just not counting what a media plan thinks it is counting.

The same page then offers a ladder. Two-second video views: the video “starts playing for at least 2 seconds in an impression session”. Six-second video views: at least six seconds, with a shorter video counting if played in full, and an engagement inside the first six seconds counting too. Then views at 25, 50, 75 and 100 percent of length. Then average play time per video view, which includes time spent replaying.

YouTube made its own position explicit this summer. From its engagement metrics documentation: “Beginning August 24, 2026, views are counted the moment a video starts to play across all formats”, including Shorts, long-form and live. For paid Shorts inventory the bar is different again, with a TrueView view counting when someone clicks or watches at least ten seconds.

So the platform hands you five or six different answers to the question did anybody see this, and the number that reaches the top of your plan is reliably the one at the bottom of the ladder, because it is the biggest. Nobody is being deceived. A delivery metric is being used as an attention metric, by us, voluntarily, every planning cycle.

What the platforms use when the money is theirs

This is the part that settles the argument, and it is available for free on two help pages.

On the same YouTube page that moves views to the start of play, one sentence survives untouched: “YPP earnings will still be based on 'engaged views' and 'engaged watch hours'.” The platform reports one currency to marketers and pays creators in another. Same video, same second, two units, and the one attached to an outgoing payment is the one with a duration in it.

TikTok goes further and prices the gap itself. Its video views objective is optimised, in its own words, for a metric called focused view, the bidding method is cost per focused view, and a focused view requires at least six seconds of play, or completion of an ad of at least six seconds, or a positive interaction within the first six seconds. Then the clause that should end every debate about this: interactions in the first second “are not counted as a focused view and are excluded from billing”.

The platform that will report a view to you at the start of play refuses to charge you for engagement in the first second, because it knows what happens in the first second is mostly accidental. It is a remarkably candid piece of product design, and it is a stronger statement about the value of a start-of-play view than any agency deck will ever make.

The platform reports a view at the start of play and refuses to bill for the first second. Both are true. Only one of them is on your media plan.

The bill nobody decomposes: 23.8 percent more for the same person

None of this would matter much if the currency were cheap and getting cheaper. It is not, and the numbers that show it are in Meta's own quarterly release.

For the second quarter of 2026 Meta reported advertising revenue of 59,363 million dollars against 46,563 million a year earlier, up 27.5 percent. It also reported three operating figures: ad impressions delivered across its apps up 14 percent, average price per ad up 12 percent, and family daily active people averaging 3.60 billion in June, up 3 percent.

Put the first and the last together, because almost nobody does. Advertising revenue per daily active person was 16.49 dollars for the quarter, against 13.32 a year earlier. That is 23.8 percent more spend to be present in front of one human being. Run it the other way as a check: price per unit up 12 percent, multiplied by units per person up 10.7 percent, which is the 14 percent impression growth divided by the 3 percent audience growth, gives 24.0 percent. Two routes, one answer.

So the audience grew three points and the invoice grew twenty-seven and a half. Twelve of those points are a straight price increase on the unit. The rest is more units pushed at the same people.

Two honest caveats before anyone quotes this at a negotiation. Meta does not break out Reels, so this is the price of feed inventory rather than of short-form specifically. And a rising price per ad is not automatically bad news; it can mean the inventory got better. The point is narrower and harder to argue with: whatever you are buying, you are buying more of it, at a higher price, aimed at an audience that is not growing. On a performance plan that can still pay. On a brand plan it is the wrong direction, because reach is what grows a brand and frequency against the same person is what grows the invoice, which is the whole argument we set out in penetration, not loyalty.

Alphabet's same quarter shows where the money is going. YouTube advertising took 11,055 million dollars against 9,796 million, up 12.9 percent. Google Network, the business of buying impressions across the open web, took 7,303 million against 7,354 million. It shrank. Feed inventory is where the budget now lives, which is precisely why its unit of account deserves more scrutiny than the industry gives it.

The six-second multiplier, and the one lever that is free

Now join the two halves. You buy starts of play. What you actually need is frames that a brand can be attached to, which means frames that are still on screen when your brand asset is visible. Take TikTok's own threshold of six seconds as the design constraint, because the platform has already told you that is where it will start charging for attention.

If your brand asset first appears at second six, only the share of starts that reach second six can do any brand work at all. Call that share your six-second rate. Your real cost per usable frame is the cost on your plan divided by that rate. The arithmetic is trivial and the consequence is not.

Your six-second rate	Real multiple of the planned cost
10 percent	10x
20 percent	5x
25 percent	4x
33 percent	3x
50 percent	2x
Rate at which the multiple is 1x	100 percent

The six-second rate is yours to measure and the rates above are illustrative. The six-second threshold, the billing rules and the view definitions are reported by the platforms.

There is no rate at which the multiple is one, so the cost on your plan is never the cost you pay. That is not a reason to stop buying short-form. It is a reason to notice which lever moves the multiple, because only one of them is cheap.

You cannot negotiate the six-second rate; it is set by the creative and the feed. You can negotiate the unit price, and you will win a few percent. Or you can move the brand asset from second eight to second one, and multiply the memory-eligible share of everything you already bought by the reciprocal of your six-second rate, at a media cost of zero. At a six-second rate of twenty percent that single creative decision is worth the same as a fivefold media increase. No media negotiation in the history of the business has delivered that.

The second lever is almost as cheap. If the platform will sell you a six-second floor, denominate the plan in it. Cost per focused view looks worse on the page than cost per view, by a factor roughly equal to the table above, and it is the only one of the two numbers that describes something a person did. A plan that gets approved on the smaller number and judged on the smaller number will keep choosing the creative that maximises starts, which is the creative that opens on a hook and arrives at the brand too late to matter.

Unbranded reach builds the category, and the leader harvests it

This is where the leak stops being waste and becomes a transfer.

An entry point is a cue that sits in a buyer's memory, and when it fires it retrieves whatever is attached to it. If your video attaches the cue to the category and not to you, the cue still gets answered. It gets answered by the brand with the broadest mental availability in that category, which in almost every category is the largest one. You paid to build the association and the market leader collects on it, because its name is what comes up when the occasion comes up.

So an unbranded short-form campaign is not neutral spending that merely fails to work. For a challenger it is a monthly payment to the leader, settled in memory rather than in cash, which is why it never appears in any review. It is also why the arithmetic of challenging a leader turns on making your move expensive to copy rather than merely visible, which is the case we set out in challenger go-to-market.

The defence is unglamorous and well evidenced: brand assets, placed early, repeated without negotiation. The useful test is not whether the logo is present but whether anything in the first second and a half is both widely recognised and uniquely yours - the two numbers that govern any asset, which we set out in distinctive beats different. On a muted phone at arm's length, a wordmark usually fails that test. A colour field, a face, a pack silhouette, a sonic cue or a signature motion usually passes.

TikTok Shop puts the entry point and the till in one place

In its seventh transparency report under the Digital Services Act, covering 1 January to 30 June 2026, TikTok states that it reaches 189 million people across the European Union every month. In the same report it notes that on 15 June 2026 it launched TikTok Shop in eight further member states: Poland, the Netherlands, Belgium, the Czech Republic, Austria, Greece, Portugal and Hungary.

The commercial read on that is obvious and mostly right. The strategic read is the one that will cost money, and it is this: when the purchase happens inside the app, the entry point and the transaction now sit in the same system, and that system will hand you an attributed sales number.

That number will reclassify short-form from a brand line into a performance line, and the reclassification will be performed by the measurement rather than by anybody making a decision. We have set out at length why a figure produced by the seller of the media is not a measurement in incrementality testing, and the specific trap here is sharper than the general one.

In-app attributed sales will be highest for the videos that show the product, because those are the videos that convert people who were already in the market this week. Those are also, reliably, the videos that cover the fewest entry points. So the measurement will reward exactly the briefs that stop building the asset, and it will be correct about this quarter while being wrong about the category. A brand that lets the shop number set the creative brief will have optimised its way into a single entry point within three planning cycles, and it will not be able to see what it lost, because what it lost was presence at occasions it no longer films.

The answer is not to refuse the commerce number. It is to refuse to let one number do two jobs. Run the shop as its own channel profit and loss on its own terms, and keep the brand case on entry point coverage with a holdout attached. Two numbers, two decisions, and no budget moving between them on the strength of an attribution window.

How we count it

Three of our lenses do the work here, and they answer different questions.

Cut-Thru-Media asks what actually gets through. We score the first second and a half of each execution on a muted phone at arm's length, and we score the asset rather than the ad, because at that length the ad has not happened yet. Shopperology asks the prior question: what are the occasions. It builds the occasion inventory for the category, scores your presence at each one, and finds the entry points the category answers and you do not, which is the list your next quarter of short-form should be built from. Brand X-Ray settles which of your assets can survive second one, on fame and uniqueness, which decides what goes in the frame before the hook has finished.

Five things to do, in order. First, write down your category entry points and score coverage per entry point rather than per channel, because channel coverage is the thing you already know and entry point coverage is the thing that grows penetration. Second, pull your six-second rate by execution and compute the multiple in the table above, so that the real cost per usable frame appears somewhere in writing for the first time. Third, move your strongest distinctive asset to the first second on everything, and measure the six-second rate again, because this is the only lever in the whole exercise that costs nothing. Fourth, re-denominate the plan in focused views, and brief against that number rather than against starts of play. Fifth, before any budget moves between the brand line and the shop line, hold out a tenth of your markets for six weeks, which we have costed elsewhere at well under one percent of an annual digital budget, and which is the only thing on this list that produces a number the seller of the media did not calculate for you.

And one thing not to do. Do not fix this by buying an attention metric and carrying on. A better exposure number is still an exposure number, and the asset you are trying to build is not seconds, it is a brand attached to an occasion. The occasion is in the brief or it is nowhere.