For the first few years I shot for the feed. Something lands well, people watch it, everyone is pleased. Then I started getting notes back from media buyers and realised I had been making a slightly different product to the one being bought.
Nothing I was doing was wrong exactly. It was just built for a place where the viewer chose to watch. An ad is served to someone who did not choose it, inside an interface that covers part of the frame, usually with the sound off, competing with whatever they were actually doing. Those are different conditions, and they change what you should ask me to shoot.
This is what I do differently now, and what I would want a brand to specify up front.
The interface eats the frame
The first thing a buyer ever pushed back on was composition. I had a habit of framing nicely — product low in the frame, face centred, a bit of breathing room at the bottom.
In an ad placement, the bottom of that frame is gone. There is a call-to-action button, a profile name, caption text, sometimes a progress bar. Whatever I put down there is sitting underneath the furniture.
So the working rule is that anything that has to be seen lives in the middle band. Product in hand at chest height rather than held low. On Reels the overlay takes roughly the top seventh and the bottom third of the frame; Stories and TikTok lose less at the bottom, but the right-hand edge carries the icon stack. I frame to the tightest of those, which is Reels, so the same file survives everywhere. It is not a creative constraint so much as a plumbing one, but if nobody mentions it at brief stage you get footage where the product is behind a button.
Do not bake the captions in
I used to burn captions into the edit, because it looks considered and it matched the styling of everything else I posted.
For ads that is actively unhelpful. The platform overlays its own text, so you end up with two sets of words fighting. Worse, baked-in captions cannot be changed — so when the buyer wants to test a different on-screen line against the same footage, or run it in another market, the file is stuck.
Now I deliver clean footage plus the caption file separately unless someone specifically asks otherwise. It costs me nothing and it leaves the asset editable at the other end.
Sound off is the default, not the edge case
A large share of the people served your ad will never hear a word of it. That is not a failure of the creative, it is just how people use their phones in public.
Which means the video has to carry meaning visually on its own. If the whole point of the first four seconds is something I say, and nobody hears it, the ad has no opening. So the product does something visible early — the pump fires, the lid comes off, the texture goes onto skin, the garment moves. Something happens that reads without audio.
The voiceover then adds to that rather than carrying it. Both layers work alone. That is the actual test.
Hooks are separate files, not one long cut
This is the single change that made the biggest difference to how useful my deliveries are, and it is a delivery-format point rather than a creative one.
Early on I would send one edit with a strong opening. If the buyer wanted a different opening, that was a new request and often a reshoot. Now, when a brief involves paid media, I shoot the body once and record the openings as separate takes, delivered as separate files against the same body.
The buyer can then assemble and test without coming back to me, which is faster for them and honestly better for me too — a test that can actually read a result needs that kind of volume, and it is far cheaper to capture in one session than to reorder in three.
Just say at brief stage that it is going into paid. It changes how I shoot the day.
Aspect ratio is a decision, not a crop
Nine-by-sixteen covers most of it. But if the same footage is going to run somewhere square or landscape, cropping a vertical video after the fact usually means cutting a head in half or losing the product entirely.
If you know a placement needs another shape, tell me and I will frame with the crop in mind — keeping the action inside a region that survives being squared off. Ten seconds of planning at my end, versus an awkward crop or a reshoot at yours.
What I still will not do
There is one place where optimising for the ad account starts working against the ad, and it is worth being straight about.
The reason this footage performs is that it does not look like an advertisement. Every step towards a polished commercial — cleaner lighting, tighter script, more takes, smoother delivery — moves it back towards the thing people have spent twenty years learning to skip. I have had briefs that asked, in effect, for a small television commercial shot on a phone, and the phone does not save it.
So I will fix the plumbing all day. Framing, clean files, separate hooks, ratios, sound-off legibility — those are free. What I will push back on is smoothing out the roughness that is doing the work, and that roughness is more deliberate than it looks.
Where the money actually leaks
Five years in, the waste I see is rarely the footage. It is the gap between what got shot and what the ad account needed, discovered after delivery, when fixing it means another booking.
Almost all of that is avoidable with two sentences in the brief: this is going into paid, and here is where it will run. Everything above follows from those two facts, and none of it costs extra if it is known on the shoot day rather than the week after.
If you have a campaign coming and the creative is going behind spend, tell me that up front and I will build the delivery around it. And if the first round works, book me again — the second one is better, partly because by then I know what your buyer does with the files.
Common Questions
How is shooting UGC for ads different from shooting for the feed?
An ad is served to someone who did not choose to watch it, inside an interface that covers part of the frame, usually with the sound off. So anything that must be seen lives in the middle band, the product does something visible in the first few seconds, and the video has to make sense with no audio at all. Feed content assumes a viewer who opted in; an ad cannot.
Should captions be burned into the video?
Not for paid. The platform overlays its own text, so baked-in captions fight with it, and they make the file impossible to re-language or re-test later. Clean footage plus a separate caption file costs nothing extra and leaves the asset editable at the buyer's end.
Why deliver hooks as separate files?
Because it lets the media buyer assemble and test different openings against the same body without coming back for a reshoot. Capturing several openings in one session is far cheaper than reordering across three, and it is what makes a creative test large enough to read.
What is the one thing to put in the brief?
That the creative is going into paid, and where it will run. Framing, delivery format, aspect ratio and sound-off legibility all follow from those two facts, and none of them cost extra if they are known on the shoot day rather than the week after.