Six Hooks, One Product: The Weekly Meta Test Grid

A performance marketer showed me his Q3 test log: 42 ad creatives shipped, 41 of them visually distinct, and not one test that told him anything. "We tried everything," he said. He had, and that was the problem. When every creative changes the hook, the visuals, the offer and the format at once, the winner teaches you nothing except that it won. He could not scale the insight because there was no insight, only an outcome.

The fix is the grid: one product, one week, six hooks, everything else held constant. It is the cheapest, most readable testing system on Meta, and AI production makes it affordable at any account size.

The six hook types

Every product gets one hook of each type. Same product footage or AI-generated environment, same body creative, same caption. Only the first three seconds change:

Why six: they cover the six reasons people stop scrolling, and one week gives each a fair test without a budget that only enterprise accounts have. Why the first three seconds only: on Meta, the hook decides the thumbstop; the body decides the conversion. Changing both at once destroys the readout.

The weekly loop

• Monday: pick the product, brief the six hooks, AI-produce the variants. Because only the opening changes, this is hours of production, not a shoot.
• Tuesday: launch all six into one ad set, campaign budget optimisation off, so each hook gets its own read.
• Tuesday to Sunday: let them run. Do not touch winners and do not kill losers before each variant clears 3,000-5,000 impressions.
• Monday: read the grid (below), keep the winner's format, brief next week's grid on a new product or a new angle on the same one.

Monthly cost at AI rates: roughly 24 finished creatives plus 72 hook variants, in the low four figures. The same grid shot traditionally would be a £20,000+ month, which is why volume testing used to be an enterprise-only game.

Reading the grid

One metric per decision, in order:

1. Hook rate (3-second views / impressions): which hook stops the scroll. The top performer is your attention message.

2. Thumbstop-to-click: which hook attracts people who actually want the product. This separates attention from intent.

3. CPA per hook: the money read. A hook can win metric 1 and lose metric 3; that combination means the format works but the message attracts browsers.

The productive pattern: when one hook type wins the money read two weeks running, that is not a hook result. It is a positioning result. A brand that finds "contrarian" winning repeatedly has learned how it should talk everywhere: landing pages, emails, organic. That is how a $100 ad test becomes a positioning decision. The volume economics that make this affordable week after week are in CAC is up, creative volume is down: the AI fix for US DTC, and the system this grid plugs into is the full Meta ad creative testing with AI playbook.

Run your first grid this week

Pick your best-selling product, brief the six hooks above, and hold everything constant except the first three seconds. The grid's whole value is its discipline; run it clean once and you will know more about your customer than a quarter of "we tried everything" ever taught.

Three mistakes that ruin a grid

The system fails in predictable ways, and all three are self-inflicted:

1. Touching it mid-flight. Pausing a hook on day two because it looks weak, or boosting a winner before the week closes, destroys the readout. The grid only produces data if every variant clears its impression threshold untouched. Judge on Monday.

2. Changing two variables. A grid where one variant has a different hook and a different body and a different offer tells you nothing — you cannot know which change moved the number. Hold the first three seconds as the only variable, ruthlessly.

3. Judging on spend instead of samples. A hook that got $8 of spend has no result yet, only noise. The 3,000–5,000 impression floor per variant is not bureaucracy; below it, hook rate is a coin flip dressed as data.

One more, subtler: running the grid in three ad sets "to compare audiences" fragments learning the same way splitting the campaign does. One campaign, six hooks, one question per week.

Reading the grid: what each outcome tells you

The value of a grid is that it produces an unambiguous next action rather than an opinion. Six outcomes cover most of what a weekly test produces.

Two of those rows are worth emphasising because they are routinely misdiagnosed.

High clicks with low add-to-cart is not a creative problem. It is the most common misread in the whole grid. The ad did its job — it earned the click — and something after the click lost the buyer. Re-editing the creative in that situation makes the numbers worse, because you lose the one thing that was working.

Flat results across six hooks usually means spend, not creative. At a small budget per variant the difference between a 14% and a 17% hook rate is inside the noise. Either raise the per-variant budget until the difference is readable, or accept that the decision will be made on the following week's data.

The five ways the grid gets polluted

A test only teaches you something if one thing changes at a time. Five things break that, and all five are common.

Changing the body as well as the hook. The most frequent error. If the body differs, a win tells you that one combination beat another — not which variable mattered.

Changing the audience mid-test. Meta will optimise into a different segment and the comparison becomes meaningless.

Running hooks for different lengths of time. The variant that ran longest accumulates the most learning. Equal budget or equal days, never a mix.

Judging on a metric the platform optimises for. If the campaign is optimised for purchases, the platform will find purchasers regardless of the hook, which flattens the hook-rate difference you are trying to read.

Stopping the test when one hook leads on day one. Early leads are noise. Three days is the minimum, and five is better for anything where the difference is small.

What to keep when the grid is done

One page per product, every week: the six hooks, the hook rate, the hold rate, the click rate, and one sentence saying what you learned. Six months of those pages is a better creative brief than any brand document, because it records what this audience actually responded to rather than what the brand believes about itself.

The six hook types worth testing

A grid only produces learning if the six hooks are genuinely different in kind rather than six phrasings of the same idea. These six cover the space, and each answers a different doubt.

Six types, one body, one product. The reason to keep the body identical is that the hook is the only variable under test — and the reason to keep the types distinct is that six variations on "buy this because it is good" teaches nothing about which doubt is blocking the sale.

One practical note on the demonstration hook: it usually wins on hold rate and loses on hook rate. That is not a failure. It means the type attracts a smaller, better-qualified audience, and whether that is good depends entirely on the cost per add-to-cart rather than on the view count.

What to do before you spend anything

Three preparation steps that cost nothing and change the outcome.

Write the six hooks as text only, and read them aloud. If the hook cannot be understood when spoken to someone who is not looking at a screen, no edit will save it.

Check that each hook survives being muted. The on-screen text has to carry the whole line, not an abbreviation of it.

Confirm the body is genuinely identical. Not "similar" — identical footage, identical captions, identical end card. Reviewers frequently let a body change slip through between variants, and that is the most common reason a clean test produces a muddy answer.