The classic direct response video formula runs roughly twelve beats: attention, problem, agitation, credibility, mechanism, proof, offer, bonuses, price justification, risk reversal, urgency, call to action. A 15 second feed ad has room for five of them. The five that survive are hook, problem, offer, one proof and the ask, with risk reversal folded into the CTA as a clause, and everything else moves to the landing page or gets cut outright.
The classic formula and the compressed one
Long-form direct response was written for someone who chose to watch. A sales letter or a 20 minute video has time to earn attention, build a case and handle objections in sequence. Feed placement inverts every assumption in that: nobody chose, nobody has sound on, and the average viewer leaves before your credibility beat.
So the compression is not "the same argument, faster". It is a different selection of beats.
| Classic beat | Share of a long-form video | Fate at 15 seconds | Where it goes instead |
|---|---|---|---|
| Attention | First 30 seconds | Kept, cut to 2.5 seconds | The hook, and it must be visual |
| Problem statement | 1 to 3 minutes | Kept, cut to one line | Second 2.5 to 5, said once |
| Agitation | 2 to 5 minutes | Cut | Implied by the shot, or the ad is for the wrong audience |
| Credibility, founder story | 2 to 4 minutes | Cut | Landing page, or a separate ad in the sequence |
| Mechanism, how it works | 3 to 6 minutes | Merged into the offer | One clause: "pulls both systems into one view" |
| Proof | 2 to 5 minutes | Kept, one item only | Second 7 to 11, shown not described |
| Offer | 1 to 2 minutes | Kept, moved forward | Second 5 to 7, before the halfway mark |
| Bonuses | 1 to 3 minutes | Cut | Landing page, where they convert |
| Price justification | 1 to 2 minutes | Cut | The price itself, stated plainly, does the work |
| Risk reversal | 30 to 60 seconds | Kept as a clause | Inside the CTA: "no card", "cancel any time" |
| Urgency | 30 seconds | Kept only if real | A date, or nothing |
| Call to action | Repeated throughout | Kept, once, in the last 4 seconds | Verb, destination, terms |
The compressed formula, written out: hook 0 to 2.5, problem 2.5 to 5, offer 5 to 7, proof 7 to 11, ask 11 to 15. That is the same slot structure we use everywhere, and how the hook, body and CTA are measured separately explains which metric holds each part accountable.
What gets cut, and why it hurts less than you think
Agitation goes first. In long-form, agitation is how you make a mild problem feel urgent to a broad audience. In feed, targeting does that job: if the viewer does not already recognise the problem from one line, the ad is reaching the wrong people and no amount of agitation fixes the buy. Agitation in a 15 second ad also reads as manipulative, because there is no room for the warmth that made it work at length.
Credibility and founder story go next. They are genuinely persuasive and they are expensive in seconds. Move them into the second ad in a retargeting sequence, where the audience has already shown interest and where a founder talking to camera does convert.
Bonuses and price justification go because they are landing-page mechanics. A stacked-value pitch needs time to accumulate, and a truncated one reads as a discount ad. Say the price instead: a real number in the first half of the ad qualifies the audience better than a paragraph explaining why the number is fair.
Urgency survives only when it is true. A stated deadline ("order before Thursday for weekend delivery") works. A permanent countdown does not, and it trains an audience to wait.
Where the proof goes when there is no time for it
Proof is the beat teams try to keep and cannot fit, so it gets replaced with adjectives. Four places to put it instead, in order of how well they work in a 15 second ad.
- In the picture. An unbroken shot of the thing working is proof, and it costs no words. A single-take screen recording, a cupboard door closing, a bag being packed. If a demo needs cuts to look fast, it is not fast yet.
- In the hook, as a number. "Month-end close took nine days" is a problem statement and a credibility marker at once, because only someone who has lived it says it that precisely.
- As behaviour, not a statistic. "I stopped buying the other three" beats "94 percent of users report improvement". It is cheaper to shoot and harder to disbelieve. What counts as social proof and what does not covers the forms that survive compression.
- On the caption card. Sound-off viewers get the proof only if it is written. One card, four words maximum.
What does not work: a testimonial montage in a 15 second ad. Three faces at two seconds each is nine seconds of your fifteen spent on people the viewer does not know, and it leaves no room for the offer.
A worked 15 second script
DTC coffee subscription. Testing a habit comparison against a price-led open.
| Slot | Timing | Line | Shot | Why it is here |
|---|---|---|---|---|
| Hook | 0 to 2.5 | "A daily flat white is twelve pounds a week." | Cafe cup on a desk, receipt beside it | Problem and credibility in one sentence, no question, a number in the first second |
| Problem | 2.5 to 5.0 | "That is the whole reason I switched, not the coffee snobbery." | Same desk, grinder arriving in frame | One line, said once. No agitation, the audience already knows this pain |
| Offer | 5.0 to 7.0 | "Six bags, fourteen pounds, cancel any time." | Box open, bags visible, price on card | Offer before the halfway mark, price spoken and shown, risk reversal already present |
| Proof | 7.0 to 11.0 | "Nobody in this office has been to the cafe since March." | Same desk in a wider shot, four mugs | Behavioural proof, one take, no statistic, no testimonial montage |
| Ask | 11.0 to 15.0 | "First box is half price this week. Tap through." | End card held: FIRST BOX HALF PRICE / TAP THROUGH | Verb, terms, real deadline. Offer repeated for viewers who joined at second 8 |
What was cut to make it fit: the roaster's story, the origin detail, the environmental angle, the comparison against two named competitors, and the free grinder bonus. All five are on the landing page. The ad's job was to get a qualified person there having already heard the price.
Running the formula as a batch
One compressed script is a hypothesis. The formula earns its keep when you run the same five slots with the offer held constant and the hook and proof varied, because then a result tells you which beat is doing the work.
That is the case for volume rather than polish: our 2026 fatigue benchmark reports CTR declining 15 to 20 percent in the first two weeks, week three landing 45 to 70 percent below launch, and brands shipping 15 to 50 variants a month seeing 3 to 5 times longer campaign lifespan than quarterly refreshers. Its conclusion is that throughput rather than talent is the bottleneck, and a fixed formula is what makes throughput possible without the batch becoming incoherent.
Genyad, our product, generates each variation as a fresh script into slots like these, using the footage you already own, with its own shot selection, voiceover, captions and export. The AI ad script generator takes the offer, the proof and the audience and writes the compressed version. Where it will not help: we have no AI avatars or synthetic presenters, so the founder-story ad in your retargeting sequence still needs a real person on camera, and we do not import from a product URL or a feed, so the offer has to be written into the brief.
Frequently asked questions
What is the direct response video formula?
Classically twelve beats: attention, problem, agitation, credibility, mechanism, proof, offer, bonuses, price justification, risk reversal, urgency and call to action. In a 15 second feed ad you keep five of them, hook, problem, offer, one proof and the ask, with risk reversal reduced to a clause in the CTA.
What should be cut first when compressing a direct response script?
Agitation, then the founder story, then bonuses and price justification. Targeting replaces agitation, and the other three are landing-page work. Cutting proof or the offer instead is the common mistake, and it produces an ad that entertains without asking.
Where does proof go in a 15 second ad?
Into the picture wherever possible: one unbroken shot of the product working, or a behavioural claim like "I stopped buying the other three". Statistics and testimonial montages need more seconds than the format has.
Does the offer really belong before the halfway mark?
Yes. Put it by second 7 of a 15 second script and repeat it in the CTA, so viewers who leave in the middle still heard the trade, and so a weak result can be attributed to the offer rather than the hook.