Mostly, not in-house. Not because it is a small job, but because it is a serious one. If you are spending money on ads, a real share of that money and a real amount of time should go into making good ones. What we always do is the read.
Creative is a core decision metric, not a deliverable at the end of a queue. Handing it to a vendor with no stake in your numbers turns a growth lever into a delivery schedule, and unless the client and the vendor are genuinely close-knit, it fails. So we are careful about when we take it on.
Creative included, owned end to end alongside everything else. That is the only shape that works, because the decisions we make on creative are the same decisions that move the account. It also means we are honest with you up front: creative demands a lot of time and a lot of investment, from us and from you.
If creative is the last box to tick, the budget line nobody wants to defend, or something to be squeezed in after the media plan is signed, we would rather say no early than take your money and produce work that cannot perform. That is not a judgement on you. It is just a bad setup, and we have seen where it ends.
Whether or not we make the work, we do the diagnosis. We tell you what your creatives actually did, gate by gate, and what that says about what to make next.
Then your in-house team takes it from there, or your creative agency does. Nobody has to take our word for the next idea. They get a brief built on evidence rather than taste, which is a much easier thing to argue with, and a much easier thing to act on.
We are not trying to win the creative line item. We are trying to make sure the money behind it is spent on the right thing.
We do not lead with angles, and we do not turn up with a fresh round of ideation and call it a strategy.
Anyone can produce fifty creative ideas in an afternoon. The hard part is knowing which of the fifty are worth making, how to run them so the result is actually readable, what the last batch already told you, and where to put more money.
A video loses people in stages, and each stage tests a different job the creative has to do. So a weak number points straight at the part that needs fixing rather than at "the creative" in general.
Share of impressions that became a 3 second view. This is the hook: first frame, opening line, pattern break.
Share of 3 second viewers who reached halfway. This tests the middle: pacing, and whether the hook's promise gets paid off fast enough.
Share of half watchers who finished. High numbers mean the offer and the end CTA are actually being seen.
The dashed line is the benchmark path: what a creative clearing all three gates looks like from impression to completed view. Every other line is one creative. Hover any line to read it, click to keep it pinned.
Share of impressions surviving each stage, on a log scale so the small numbers at the end stay readable. Each line is coloured by the gate it misses hardest, so the shape of the account is visible before you read a single number.
Illustrative. This is the benchmark path with a representative spread of creatives, not a client account.
Two videos, same spend, same offer. One hooks twice as many people and still delivers less. If the thumb stop ratio is all you look at, you kill the wrong one.
Both lines are indexed to 100 at the 3 second mark, so this reads as: of the people you did hook, how many stayed to the end.
| Progress | Creative A | Creative B |
|---|---|---|
| 3s | 100 | 100 |
| 25% | 50 | 82 |
| 50% | 21 | 64 |
| 75% | 13 | 55 |
| 100% | 8 | 50 |
Illustrative, drawn to a shape we see repeatedly. Creative A wins on the metric most decks report and loses on the one that pays.
A hooks twice as well and still loses. It buys the stop with a first frame that has nothing to do with the offer, so the people it stops leave the moment they work that out. B stops fewer people and keeps half of them all the way to the CTA.
Cut A and you have learned nothing. Put A's hook on B's body and you have a different business.
Retention tells you where people left. The work is understanding why the rest stayed, because that is the only part you can deliberately repeat.
The first frame, the opening line, the pattern break. Cheap to test, and the easiest thing to mistake for the whole creative.
When completion is high and the hook is ordinary, the offer is usually doing the work. That changes what you make next, and what you brief your studio to stop doing.
A face, a format, a length, a voice, an unglamorous product demo. Often the thing nobody pitched. You only find it by looking at the ones that finished.
If nobody is asking for these numbers, you are paying for design aesthetics, not results.
That is the whole test, and it applies to us as much as anyone. Whoever makes your performance creative should be asking for thumb stop, hold and completion before they ask for a moodboard. Because if they cannot tell you what worked last month, they are throwing darts in the dark next month and waiting to see what sticks.
We would rather read your last three months than pitch you fifty new ideas. You can hand the read to your own team or your own agency, and we are glad when you do.