Most brands book a content shoot to get three great ads. They come back with three great ads, run them, watch performance decay after two weeks, and have nothing left to test. The production was competent. The brief was wrong.
A shoot day should produce ingredients, not ads. Nobody should be trying to capture a finished ad in camera, because a finished ad is an assembly job and assembly happens in the edit, cheaply, dozens of times over.
I ran a production day recently for AURÊLO (opens in new tab), the Alcantara iPhone case brand I co-founded with four partners and run creative for. This is how we structured it and what came out the other side. It's a playbook, not a theory, so it comes with the boring parts included.
Two deliverables, not one
Everything on the day serves one of two outputs.
A clip library. Between 100 and 150 modular clips, three to eight seconds each, shot vertical in 4K, with nothing baked in. No on-screen text, no music, no call to action, no hard cuts you can't undo. Every clip has to be able to sit in an ad you haven't thought of yet.
A voice bank. Talking-head takes plus clean voice-over lines read to mic, per creator. Voice-over is the highest-leverage thing on the entire day because it can be married to footage shot months later, and it costs almost nothing to capture while a lighting change is happening.
If you finish the day with those two assets properly logged, the day worked. If you finish with six polished films, it didn't, no matter how good they look.
Hook, body, CTA
This is the mental model that makes the rest of it make sense. Every finished ad is three interchangeable parts.
Hook, zero to three seconds. On-screen text, added in the edit. Never spoken on set. Hooks are the thing you'll iterate most and the thing that fails most often, so they must never be welded to footage.
Body, fifteen to twenty-five seconds. The footage and the creator's voice. This is the only part of the ad the shoot day exists to capture.
CTA, the tail. Shop now, shipping promise, "match your fit", the URL. Added in the edit.
Now the multiplication. One body, fronted with five hooks and closed with two CTAs, is ten distinct ads with no reshoot. Ten bodies handled the same way is a hundred. Volume and variety on the day beat polish on the day, because polish improves one ad while variety improves the entire test surface. Same argument I made about building in The Sunday Test: the cost of the raw material has collapsed, and the constraint has moved to knowing what to do with it.
How to structure the day
Split the day into blocks by content type, and put them in priority order so that when you run late, which you will, you cut from the bottom rather than losing the material you can't replace.
1. Automotive and hero environment, about 25 per cent. Arrival, movement, environment. This is where the production value lives and it's the hardest to recreate on a phone next Tuesday, so it goes first while the light and the location are yours.
2. Micro and product detail, about 25 per cent. Texture, material, mechanism, close-ups. Shoot 30 to 50 clips in this block alone. It's the cheapest content to capture and the highest reuse on the account, because detail shots are model-agnostic and colour-agnostic. They drop into every variant you'll ever cut. Under-shooting this block is the most common and most expensive mistake of the day.
3. Lifestyle, about 25 per cent. The product in real context, styled, in use, in someone's actual hand and pocket.
4. Talking-head UGC, about 25 per cent, and first to be sacrificed. Not because it doesn't matter but because it's the only block you can genuinely recreate later without a production day. A creator can film a talking head in their kitchen. They cannot film your hero environment in their kitchen. And the voice bank doesn't need this block at all: capture voice lines in the gaps, between setups, while someone's changing a lens.
One tactic that pays for itself every time: change wardrobe between blocks. Swapping even a single item makes the same creator on the same day read as a different shoot. It is the cheapest variety multiplier in production and it costs you ninety seconds.
Four rules for anyone speaking on camera
Brief this before you roll, not after you've wasted a take.
- Say the real material name. Alcantara, not "suede". Getting this wrong doesn't just weaken the ad, it misdescribes the product.
- Never invent a spec or a claim. Only what's scripted or personally true. A creator improvising a warranty you don't offer is a problem you now own.
- No prices on camera. They change, and a price-locked clip has a shelf life. Use durable claims instead: shipping speed, guarantee, fit.
- Get the numbers right and consistent. Exact count of colourways and models. A creator confidently saying the wrong number turns good footage into unusable footage, and you won't notice until the edit.
The marque rule
If your product borrows prestige by association, say a material used in premium automotive interiors, there's a distinction worth understanding: naming a marque spoken, inside natural testimonial audio carries a different exposure profile to the same name burned into an on-screen text hook or the ad copy.
Text reads as a brand claim. Speech reads as a person's observation. Implied endorsement and IP complaints tend to follow the former far more readily than the latter, and takedowns land on ads, not on interviews.
So the working rule we use: association can be spoken, sparingly and truthfully, by a person describing their own experience. Association does not go in the hook, the headline, or the primary text. I'm an operator, not a lawyer, and anything genuinely borderline should go past one. But this rule has kept a lot of creative live that would otherwise have been pulled.
Turning the day into fifty ads
Log and tag every clip immediately. Three axes: content type, creator, variant or colour. Do it the same week or it doesn't happen. An untagged clip is a dead clip, because at 150 files nobody is scrubbing through footage to find the one where the light was right.
Build bodies before you build ads. Aim for eight to fifteen bodies, in three types, all pulling from the same library:
- Talking head. Continuous take, B-roll cut over the middle third, back to face, into the CTA.
- Voice-over on B-roll. Your best one-liners as audio only, no face on screen. Cheapest to cut, endlessly re-editable.
- Silent. B-roll, one line of text, music. Reads as premium, needs no voice, and often outperforms the ones you spent the most on.
Then multiply. Front each body with different hooks, close with different CTAs. Eight to fifteen bodies becomes thirty to fifty variants.
Name files so results are readable. Something like creator_bodytype_hook_variant. This is not admin. It's the only way you'll be able to answer "what's actually winning" a fortnight from now, and the answer is never "the video" — it's a hook, a body type, or a creator.
Load broad, not narrow. Few campaigns, many ads inside each ad set, broad targeting. Let the system allocate to winners instead of pre-deciding them for it. Pre-segmentation is you assuming you know the answer before the test has run.
Read two signals early. Hook rate, meaning retention past three seconds, tests the hook. Cost per purchase tests whether the thing sells. They fail independently, and that's the useful part: a strong hook rate with poor conversion means the front is working and the body isn't. Recombine the winners, cut the losers, and keep the library open.
Why this is a brand problem, not a media problem
Media buying isn't my lane. Creative systems are, and this is where most brands actually lose: not in the account, but in a production brief that guaranteed they'd run out of things to test.
Creative fatigue is a supply problem. One well-run day, structured as ingredients, feeds weeks of testing. The same day structured as "let's get three great ads" feeds a fortnight and then you're paying for another shoot, at which point the production cost per ad is roughly ten times what it needed to be.
Get the brief right and the day gets cheap. That's the whole trick, and it's the same trick as everything else I write about here — see The Founder Stack for the version that applies to running companies rather than shoots.
If you're briefing a shoot, rebuilding a brand's creative supply, or working out why competent production isn't producing results, that's the brand and positioning side of what I do.