Problem-solution video ads: a second-by-second beat sheet for 15 and 30 seconds

What do the extra 15 seconds of a 30-second cut buy you?

They buy one consequence and a longer demonstration. The opening stays the same: problem by 0:00, proposition by 0:03.

The consequence beat answers why the problem matters today, and in the example the stained cuff ends up on camera in a video call. Google's ABCDs of effective video ads call this Connection, "Help people think or feel something", and the beat earns its four seconds only if the viewer recognises the moment from their own week.

The remaining time goes to the mechanism, which grows from five to nine seconds, and to a proof beat long enough to show the before and after on the same fabric. Spend nothing extra on the logo. On AI Vidia's Performance Retainer every video runs up to 30 seconds in 9:16, so both cuts in the timeline fit one brief.

Where should the brand appear in a problem-solution ad?

Early, small, off centre and more than once. Google's ABCD guidance says "Brand early, often, and richly" and asks brands to "Show up early and throughout". Google also quotes a Kantar study (April 2021) crediting ads that follow the ABCDs with as much as a 30% lift in short-term sales likelihood; it is a Google and Kantar study of Google's own guidance, so read it as a vendor study.

The stronger evidence on placement is older and comes from television. Teixeira, Wedel and Pieters (Marketing Science, 2010) combined eye tracking and zapping data from nearly 2,000 participants watching 31 commercials. Brands placed in the centre of the screen increased avoidance, while brand size did not. Pulsing the brand in short appearances reduced avoidance at the same total brand exposure.

Those were TV ads in 2010, watched with a remote, and a feed swipe is a faster exit. Our reading of both sources together, which is an inference: put a small brand cue near the edge of the frame by 0:02, bring it back during the demonstration, and close on the end card. The timeline marks those pulses at 0:02, 0:12 and 0:26 in the 30 second cut.

How far can you dramatize the problem?

Exaggerate the annoyance, never the result. A spill can be bigger than life; the clean fabric afterwards has to be what the product actually does.

Two rulebooks set most of the limits. Meta's health and wellness ad standards prohibit ads that contain "statements of inferiority about physical appearance" and "promises of specific outcomes within a set timeframe without disclaimers or qualifiers". The FTC's Health Products Compliance Guidance (December 2022) says the law looks at "the consumer's understanding", so images and demonstrations make claims too, health claims need "competent and reliable scientific evidence", and testimonials "don't constitute substantiation".

Generated footage adds another trap. The UK ASA and CAP warned in Generative AI and advertising (31 August 2023) that AI images used to make efficacy claims have "the potential for them to mislead", and that advertisers keep "the primary responsibility" for compliance.

In practice the consequence beat may embarrass the shirt. It should never mock the person wearing it.

Which beats can AI generate, and which need the real product?

Generate the problem and the setting; take the product and the proof from the real thing. The table pairs each beat with its timing in both cuts, the source behind it and the claim check to run before the cut goes out.

The split follows the failure modes in which products are hard for AI video: models reproduce a setting more reliably than a hand pressing a pump or text on a label. The AI video ad production service builds each cut from your product photos and a signed product fact sheet for exactly that reason.

If the voiceover carries the problem, write the speech track and the picture track separately, as the AI UGC script framework shows, so the ad still works muted.

How should you test a problem-solution ad against your current best?

Test problems, not edits. Two cuts that open on different problems are two concepts, because they speak to different buying moments; two cuts that differ only in the end card are variants of one concept and teach you far less.

Keep the mechanism and proof beats identical across cells and change the problem shot and the proposition line. Then run the new concepts against the ad you already trust, as in testing new AI video against your best ad. If you want that run as a fixed, measured comparison, the Control Test sets up the cells and the read.

Start with one product today.

Take your best seller and spend 20 minutes on three boxes: the problem a stranger would recognise in a single frame, a proposition of six words or fewer, and the proof you can show on the real product. If the proof box stays empty after 20 minutes, brief a demonstration ad instead and come back to this format once you have a result you can film.