Research insights · Descriptive editorial pilot
Brand and message in 18 opening-frame samples
An opening can explain something without naming the advertiser. A brand can be visible without explaining the offer. We inspected two fixed moments in nine public ad examples to keep those questions separate.
By Oussama Nakhil · Updated 8 October 2026
The observation, with its limits
Of 18 sampled frames, seven had a clearly readable advertiser name, nine did not and two were uncertain. Three contained explanatory or offer text; two of those also contained a readable advertiser name. These are frame counts from a convenience sample, not percentages of effective ads or estimates of what viewers noticed.
The observations were coded with AI assistance by Codex. There was no independent human rater, consumer sample or campaign test. SaliencyLab publishes this analysis and sells creative-review services; this is not evidence that its predictions improve advertising results.
Three editing questions the samples expose
- Does the explanation identify its source? At 2.0 seconds, the CeraVe sample explains the skin-barrier metaphor but has no readable advertiser name. Test whether an earlier name helps attribution while preserving the explanation. Familiar colors may already cue the brand; the frame alone cannot settle that.
- Is the name still legible after the cut? The BMW Meta sample has readable badge lettering at 0.5 seconds. At 2.0 seconds the badge remains visible, but its lettering is uncertain. Check recognition on the actual playback size rather than equating a visible badge with a readable name.
- Does the claim work at viewing size? Booking.com combines its name with booking copy at 2.0 seconds. Nivea combines its name with a 72H pack claim. Our native-resolution check does not prove either claim is readable on a phone, understood or believed.
A practical next step is to review the full opening with sound on and off, identify the intended cue and then compare one controlled edit. Avoid adding a logo or caption automatically when it would obscure the demonstration.
Sample and reproducible method
We included all nine videos linked by the public case-study archive at the time of this check. We extracted frames at exactly 0.5 and 2.0 seconds with FFmpeg and inspected them at native resolution. A readable name can be on packaging or a badge. Brand colors, a familiar silhouette and an unreadable mark do not count as a readable name. Explanatory or offer text communicates more than the brand or model name alone; a readable duration claim on packaging counts.
The protocol was written before this new coding pass, but the archive and its prior model labels were already known. This was neither blinded nor preregistered. Two samples cannot establish the first brand appearance, continuous first-three-second coverage, sound-off comprehension, attention or recall. Archive filenames identify the assets, not independently verified media placements or market targeting.
Download all observations and source hashes (JSON). Download the extraction and coding method. Third-party videos retain their original rights; these downloads contain observations and links, not a licence to reuse the ads.
Every observation
| Asset / time | Readable name | Explanatory text | Visible evidence |
|---|---|---|---|
| 03-cerave-tiktok-uk · 0.5s | no | no | Presenter and wall demonstration; no readable advertiser name. |
| 03-cerave-tiktok-uk · 2s | no | yes | BRICK WALL = SKIN BARRIER explains the demonstration; no readable advertiser name. |
| 05-loreal-paris-meta-us · 0.5s | no | no | Woman in a lift; no readable advertiser name or explanatory copy. |
| 05-loreal-paris-meta-us · 2s | no | no | Lobby scene and partial background signage; no readable advertiser name or explanatory copy. |
| 06-coca-cola-meta-us · 0.5s | unknown | no | Red bottle label is motion-blurred. Familiar packaging alone does not satisfy readable-name coding. |
| 06-coca-cola-meta-us · 2s | no | no | Industrial structure; no readable advertiser name or explanatory copy. |
| 07-bmw-youtube-de · 0.5s | no | no | Moving vehicle; badge too small to read an advertiser name. |
| 07-bmw-youtube-de · 2s | no | no | Distant moving vehicle; no readable advertiser name or explanatory copy. |
| 08-booking-youtube-ae · 0.5s | yes | no | Booking.com name over palm trees. |
| 08-booking-youtube-ae · 2s | yes | yes | Booking.com and Reserva l’estada ideal appear together. |
| 09-nivea-youtube-de · 0.5s | yes | no | NIVEA name in a blue circle. |
| 09-nivea-youtube-de · 2s | yes | yes | NIVEA name and a readable 72H claim on the deodorant pack. |
| 10-bmw-meta-de · 0.5s | yes | no | BMW lettering readable on the steering-wheel badge. |
| 10-bmw-meta-de · 2s | unknown | no | Hood badge visible, but lettering not clearly readable at native resolution. |
| 11-nike-youtube-us · 0.5s | yes | no | NIKE INVINCIBLE 3 names brand/model; model name alone is not explanatory or offer copy. |
| 11-nike-youtube-us · 2s | yes | no | NIKE INVINCIBLE 3 remains on screen; no explanatory or offer copy. |
| 12-netflix-meta-us · 0.5s | no | no | Character in a trailer scene; no readable advertiser name or explanatory copy. |
| 12-netflix-meta-us · 2s | no | no | Truck scene; no readable advertiser name or explanatory copy. |
What would make this stronger?
An independent second coder could check the same frames, report disagreements and revise the coding rules before a larger sample. A playback study with real target viewers would be needed to test recognition and comprehension. A campaign experiment would be needed to estimate an outcome effect. We have not run those studies.
For a fuller worked example, see the CeraVe diagnosis and controlled test plan. For choosing the evidence appropriate to your decision, see the creative-testing method comparison.