How to test ad hooks

To test hooks, write several openings for one ad, keep everything after the opening identical, give each variant comparable delivery and judge them on a metric agreed before launch. Hook rate shows which opening stops the scroll. The business metric, such as CPA or ROAS, shows which one is worth scaling, and the two don't always agree.

What counts as the hook

The hook is the opening of the ad, the part a viewer sees before deciding whether to stay. It has three layers that work together: the first frame, the first spoken line and the on-screen text. A hook test can change one layer or all three, as long as the test plan says which.

TikTok's own guidance gives a sense of the window. It recommends introducing the content proposition in the first 3 seconds and placing the hook within the first 6 (TikTok, June 2025).

Write hooks from different angles

Hooks that differ by a word or two rarely produce a difference you can read. Hooks that come from different angles do. For each concept we write across angles like these.

Keep the body constant

A hook test only works if the hook is the only difference. We film one body per concept and edit each hook onto it, so every variant has the same demo, the same offer, the same end card and the same length after the opening. If two hooks need different bodies to make sense, they are two concepts and belong in a concept test.

The metrics

We read hook tests on four numbers. The definitions below are ours, built from metrics that every platform reports under its own names.

Give each variant a fair run

When variants share one budget, the delivery system can favor an early leader before the others have had a fair run. We set a minimum spend per variant in the plan, expressed as a multiple of the target CPA, and structure campaigns so each hook reaches it before anything is judged. Variants that miss their threshold on the decision metric are cut, and the rest move to the next round.

How many hooks a budget can test

The number of hooks follows from the budget, not the other way round. Multiply the spend threshold per variant by the number of variants, and the total has to fit inside the test budget with room to spare for delivery that doesn't split evenly.

When the budget can't cover the matrix, cut creators before cutting hooks in a hook test, and cut hooks before creators in a creator test. A smaller matrix where every cell reaches its threshold gives a result you can act on, which a wide one spread thin rarely does.

Hooks and the placement

The same opening can behave differently by placement, because the interface and the ad format around it differ. On YouTube Shorts, for example, the call-to-action appears 3 seconds into the ad for Performance Max, App and Demand Gen campaigns (Google, read 6 Oct 2026), so the opening shares the screen with it sooner than on some other placements. We read hook tests per placement before pooling them.

Read past hook rate

A hook that shocks or teases can post the best hook rate in the test and the worst CPA, because it stops the wrong people or promises something the body doesn't deliver. So hook rate alone never decides a test. The pattern we look for is a hook that lifts hook rate and keeps that lift through hold rate and click-through, all the way to the decision metric.

When the numbers split, hold rate usually explains it. A high hook rate with a weak hold rate means the opening and the body don't fit, which is a problem for the next script as well as the next hook.

Iterate from the winner

• Put the winning hook on a new body or a new creator, to see whether the angle carries.
• Write new hooks in the winning angle, to find a stronger version of it.
• Retire angles that lose twice, and record them in the test log so nobody retries them by accident.

On-screen text is part of the hook, so it has to sit where the interface won't cover it. Meta suggests leaving at least 14% of the top and 35% of the bottom of a Reels ad clear (Meta, read 6 Oct 2026), and Google gives 10% at the top, 25% at the bottom and 10% on the right for Shorts (Google, read 6 Oct 2026). TikTok also recommends showing 5 to 10 words per second when using text (TikTok, June 2025). Our spec sheet has the full list.