About a third of the ads we make are filmed by someone who is not us: a customer, or a creator who uses the product. When these films work they outperform anything from our studio. When they fail, they fail in the same way every time: a stranger holding a product at arm's length and reciting our adjectives back to us.
The difference is almost entirely in the brief. Here is how ours changed.
We stopped sending scripts
Our early briefs were two pages long and contained a script. People read it to camera, word for word, with the slightly hunted look of someone doing a favour. Across the 90 customer films we commissioned in 2025, the scripted ones held 22% of viewers to three seconds. The ones made after we dropped the script held 35%.
A script takes away the only thing a customer has that we do not, which is their own way of saying it.
The brief is now three questions
We send a short voice note and three questions. What were you using before, and what annoyed you about it? When did you first notice this was different? Who would you tell about it, and what would you say? We ask people to answer out loud, in the place where they actually use the product, and to show it in use while they talk.
Then we add four practical requests. Film upright. Stand facing a window. Start talking in the first second, with no greeting. Send us everything, including the takes you think are bad.
The takes people apologise for are usually the ones we use.
We pay for footage, not for a finished ad
Customers are not editors and should not have to be. We ask for five to ten minutes of raw material and cut it ourselves, often into three or four different ads. One long ramble about sunglasses on a ferry gave us a fifteen-second story, two openings for other films and a line we later used as a headline.
The Sundry Eyewear story shown here began as that ferry footage. The headline is the customer’s sentence with one word removed. It ran from April to August 2026 inside an account that returned 3.4x on ad spend over the same period.
Everyone who films for us signs a one-page agreement in plain language: what we may use, where, for how long, and what they are paid. Twelve months of use is our standard. If an ad is still working after that, we go back and pay again.
Choosing who to ask
Follower counts are close to irrelevant, because the ad will be shown to strangers by a media buyer, not to the creator’s own audience. We look for two things instead: someone who already talks with their hands, and someone who has really bought the product more than once. Repeat customers tell better stories, because they have had time to form a habit and can describe it.
We find most of them in the reviews. A person who has written four specific sentences about a face serum without being asked will usually say six good ones on camera.
What we still get wrong
Roughly one film in four is unusable, and we have not found a way to predict which. The light is wrong, the nerves show, or the story turns out to be about delivery rather than the product. We budget for that rather than fight it: for every three customer ads in the monthly plan we commission four films.
What we no longer do is try to rescue a weak film with graphics and music. If the person is not believable in the first two seconds, nothing we add in the edit will make them so. Better to thank them, pay them, and ask the next customer the same three questions.




