What the numbers show when creative is managed weekly

Every paid social account already contains the most useful number on this page, and almost nobody computes it. Put several ads behind one media plan, same budget, same audience settings, same dates, and they will not perform the same, the share of people who watch the best will be several times the share who watch the worst, and that gap, the spread, is the size of the opportunity a weekly loop has to work with. What follows is the spread as the platform itself publishes it, then what we have seen in named and anonymised accounts when a portfolio is optimised weekly against it. These are measurements, in the accounts they came from, against each brand's own existing creative. They are not a promise, because the next account's spread is its own.

The spread

Meta's own test data puts the industry-level number on it, from the platform that sells the media, so read as such. Across 235 brand lift tests run between April 2021 and 2022, the winning creative delivered a 13% lower cost per ad recall than the average of the other creatives in the same test. Across 14,780 A/B creative tests between April 2021 and April 2022 it delivered a 21% lower cost per website conversion and a 27% lower cost per app install, again against the average of the other ads in the same test. Same media, same audiences, same dates, different work.

Two things about those figures matter more than their size. The comparison is winner against the average of the rest, not winner against worst, so the full distance between the ends of any one test is wider than the number shows. And the more distinct ads sit under one plan, the further apart the ends sit, so a portfolio of twenty to thirty distinct concepts shows a wider spread than a split test of a few variants ever will. The spread is the structure of the medium rather than an anomaly to be ironed out, because the auction prices every impression partly on how people respond to the ad and the delivery system matches each ad to the people who respond to it. Why creative is your largest working media lever explains that mechanism. The consequence is the point of this page. Under a fixed media plan the account's result is a weighted average of its ads, and every week the weights can be moved.

What a weekly loop does with it

A weekly loop is the operating answer to the spread. Each week the portfolio is optimised. The ads that are spending without being watched are paused, the ads that are being watched are extended, and what the watched ads have in common, the opening, the person, the situation, becomes the next brief. An ad is retired because it was not watched, never because it is old. Over three or four cycles the spend migrates from the bottom of the spread to the top, which is why the account-level numbers below move more than any single ad did, and why they are read against the brand's existing creative on the same plan rather than against a benchmark.

One measured account

In one measured account, same budget, same media plan and targeting, across three weekly cycles, the brand's existing creative was buying a completed view at $0.50 and the social-first portfolio reached $0.09, with a completion rate five times as high as the existing creative's. Part of that gain is the loop retiring weak work, each week's pause decisions removed the ads that were spending without being watched, and the rest is the auction rewarding the ads that were watched with cheaper delivery. Both are creative levers. The media buying never changed. One account, so it illustrates the mechanism rather than proving a law, and the decomposition is what to carry. Most of the movement is the completion rate itself, more of every thousand impressions becoming a watched ad, and the remainder is the price of those impressions falling because the platform preferred to show them.

The media bill

A large global CPG company showed the media half on its own. On the same media plan the portfolio cleared at a 48% lower CPM than the brand's existing benchmark, before counting a higher completed view rate on what was delivered. CPM is the media team's number with a creative contribution, and the direction of that contribution depends on the objective, under reach and awareness objectives more relevant creative clears cheaper, under conversion objectives it clears dearer and lowers the cost per outcome, which is why the standard in How to read paid social results states the objective beside every efficiency figure. Nobody negotiated that reduction. The auction priced the work, impression by impression.

What moved for the brand

Performance and efficiency are the mechanism. Effectiveness is the verdict, and the standard reads it first and on its own clock, because a cheap watched view that did not move the brand is cost-effective waste and the standard says so. Three brands we can name. Vaseline saw three times the brand lift of the brand's existing creative against the same brief and media spend. Whiskas saw five times the brand favourability lift of its existing creative on the same media. KitKat, a brand everyone already knew, saw two times the purchase intent of its existing creative in lift testing against the same media. Each multiple is against the brand's own prior work in the brand's own account, which is the only baseline we lead with, because the hook and hold benchmarks circulating online largely cite each other. And each is a lift result, so it says the brand moved, which is a different claim from the ads being watched, and the performance tier had already made that one.

The Advertising Research Foundation's 2026 validation work is the right frame for the whole page. Attention translates into outcomes alongside media weight, reach and frequency, so none of these numbers replaces the media plan, they describe what the media plan delivered once the creative under it was managed.

Show me sales

The strongest objection to everything above comes from finance and from the effectiveness purists, and it is fair. These are platform numbers and vendor numbers. Completed views are what Meta counts, brand lift is what a platform or a research vendor measures, a brand pays its shareholders in sales, so show me sales. Three answers, in the order the standard reads results.

First, sales lift and marketing mix modelling are the brand's instruments, run across everything the brand does, and where a client runs them we report what they show and we do not produce them, because a creative partner claiming a sales number it did not measure is borrowing a result. Second, the lift studies are the effectiveness tier we have, and they sit above every performance number on this page for a reason. They say the brand moved, with the method named, in-market against a holdout or forced exposure, because the two answer different questions and a platform running its own lift study is grading its own homework, which is worth saying out loud. A result is only called a result when the tiers are read together. Third, the performance numbers are not the claim, they are the mechanism, and they are the brand's own data, pulled from the brand's own account. So we claim the performance tier at whatever media plan the brand chose, a contribution to media efficiency where it was measured with the objective stated, effectiveness only where a study read it, and no universal multiplier.

Media creates the opportunity to see your ads. Creative determines whether they work.

What to do with this

Compute your own spread. Pull every ad from the last quarter, one platform, one threshold, Meta ThruPlays over impressions or TikTok focused views over impressions and never the two blended, and rank by watched rate. Mark the brand's existing hero creative on the list. The distance between the top ad and the bottom ad is the opportunity, and the distance between the top ad and the account average is roughly what a loop that moved spend weekly would have captured. Then ask how many of the bottom third were still spending at quarter end, because the spread is only an opportunity if someone acts on it every week. How to audit your own paid social account turns this into a worksheet. Then read How to work out what your creative really costs, which puts the creative fee into the same arithmetic and gives procurement the one number that can come out for or against it.

Questions

Will we see the same results?

Nobody can promise that and we do not. Every number on this page is a measurement in a specific account against that brand's own existing creative on the same media, and the next account's spread is its own. What transfers is the mechanism. Under one media plan, distinct ads perform differently, the auction prices them differently, and a weekly loop moves spend from the ads that are skipped to the ads that are watched.

The size of the gain depends on the size of the spread in your account and on how good the existing creative already is, which is why the first thing to compute is your own spread, not ours.

Why report against our own creative rather than a benchmark?

Because the benchmark hook and hold rates circulating online largely cite each other rather than any primary sample, and because a benchmark cannot hold media constant. Your existing creative, in your account, on your media plan, is the only comparison where the media is the same and the work is the only variable.

It also keeps the arithmetic honest in both directions. If a portfolio does not beat the brand's existing creative on watched rate and all-in cost per watched impression, that comparison says so, and the standard requires that we report it.

How quickly do the numbers move?

Performance moves within weeks. In the one measured account the cost per completed view moved across three weekly cycles, because each cycle retired the ads that were not being watched and extended the ones that were. Media efficiency moves with it, since the auction reprices the ads as the response to them changes.

Effectiveness is read on its own clock. A lift study needs a minimum spend and a period to run over, so the brand verdict arrives after the performance picture is already clear, and we do not borrow it from the performance tier in the meantime.

Sources

• Meta, "Creative and Media Excellence", circa 2023. Winning creative against the average of the other creatives in the same test, 235 brand lift tests (April 2021 to 2022) and 14,780 A/B creative tests (April 2021 to April 2022). Meta internal test data, platform interest, Meta sells the media.
• Meta Business Help Centre, "About brand lift tests". Exposed and holdout methodology. Platform documentation, Meta runs its own lift studies.
• Advertising Research Foundation, Attention Measurement Validation Initiative Phase 3, press release, 9 June 2026. Attention translates into outcomes alongside media weight, reach and frequency. Member-funded industry body.
• One measured account, same budget, same media plan and targeting, anonymised. Cost per completed view from $0.50 to $0.09 across three weekly cycles, completion rate five times as high, part of the gain from the loop retiring weak work. Our own account data.
• A large global CPG company, anonymised. CPM 48% below the brand's existing benchmark on the same media plan. Our own account data.
• Vaseline, Whiskas and KitKat, lift testing against each brand's existing creative on the same brief and media. Cleared proof points from our own accounts.