Media is bought to the cent. Creative arrives by the shoot. One tested creative lands near $1.80.

The test gets written properly. Three hooks, two offers, two kinds of proof. Twelve cells, each with a question attached to it.

What arrives is three files. Usually the three that were easiest to shoot on the day.

So the account learns which of three worked. The other nine questions stay open, and they were the interesting ones.

Nobody rejected them. They were dropped in a scheduling conversation, by people who were not in the room when the test was designed.

Then the report says creative underperformed. The honest sentence is that creative was under sampled, and those are different problems with different fixes.

The wrong one gets funded, because another round of audience work is easier to approve mid-flight than another round of production.

Every buyer knows this. Few can say it in a review, because the creative budget was signed off by somebody more senior than the person watching it fail.

Published benchmarks put the share of ad creatives that become winners at roughly 5 to 8 percent. The figure comes from Motion's analysis of 550,000+ Meta ads.

Call it one in thirteen to one in twenty.

Now hold three creatives a month against that. The expected number of winners is well under one, and the month still gets reviewed as though the media plan were the variable.

Volume is not a creative philosophy. It is the sample size the win rate demands, and it is the one input on the plan nobody costed.

The lever is not the win rate. It is what one attempt costs, because that is what decides how many attempts get made.

There is a second effect, and it is the one buyers feel on a Wednesday. An ad that cost six hundred dollars to produce has a defender in the room, so it runs until Friday. An attempt that costs under two dollars gets switched off the moment the hook rate says so.

The blocker is the shoot, so remove the shoot.

An account manager answers the same objection eleven times a week. That answer, said once into a phone at a desk, is a creative.

A customer sends a voice note about why they switched. Ask them for forty seconds of it to camera.

A founder explains the thing the landing page buries in paragraph four.

None of that needs a studio, a light kit or a booking. It needs three minutes and a quiet room.

Cutroom takes the take, transcribes it, drafts the cut, places b-roll on the words and returns a finished vertical MP4 with captions burned in.

You direct it on the transcript. Highlight a phrase and a clip lands over exactly those words. Delete a weak line and the cut rebuilds around it.

A variant stops being an afternoon. It becomes a decision about the ad, taken on the same screen where you read it.

A batch is 100 credits and covers the transcript, the director pass and the b-roll search. Rendering is 20 credits per output minute, so a thirty second ad adds 10.

Call it 110 credits an attempt.

Twelve cells is about 1,320 credits. Basic is $39.99 a month for 2,500 credits, so the whole matrix runs with change left for the two cells you want to redo.

Lite is $19.99 for 1,250 credits, about eleven attempts. Premium is $79.99 for 5,000 credits, about forty five.

Per attempt that is roughly $1.80 on any of the three plans, which is the first time the creative line on the plan has had a unit price at all.

A different opening headline and a different caption style export from the same batch at 20 credits an output minute, so the second version of a cell costs the render and nothing more.

Run that back through the first section. The nine cells that were dropped in a scheduling conversation now cost about sixteen dollars between them.

Every export is a 1080x1920 vertical MP4, the shape Reels, TikTok and Shorts all want. There are no landscape masters and no square crops, so a placement plan that needs a 16:9 master needs a second route for that placement.

There are no subtitle files to download either. Captions are burned into the picture, which is what a sound off feed reads anyway.

Source takes cap at three minutes, and one take makes one finished ad.

Somebody has to speak. If nobody will, the presenter module turns a photo and a voice-over into a lip-synced face at 350 credits per minute of render, and generated voice-over is 70 credits a minute.

Coverage is capped so the ad still reads as a person talking: 40 percent of the runtime under b-roll at a chill pace, 45 at normal, 52 at fast.

That cap is a performance decision rather than a limitation. Cold traffic responds to a person making a claim, and a montage with a voice over the top is the thing they scroll past.