643,339 views. 231 likes. Here's what the first four seconds did.
Ralph Lauren's Wimbledon film is a genuinely lovely piece of work. We scored it second by second with 179 simulated viewers, then put the result next to the channel's own numbers. The creative is not the problem.
Above: every beat of the film, scored by all 179 simulated viewers. Brain load is a derived processing-effort index, not a neural measurement.
The film
"Wimbledon's Chief Inspectors of Excellence" runs 143 seconds. Two children, Lily and Oliver, are appointed Ralph Lauren Chief Inspectors and sent to audit the tournament: net height, grass length, colour swatches, the strawberries, the ball boys' uniforms, and a resident hawk called Rufus. It is charming, well made, and it earns its jokes.
We ran it through the simulation layer: 179 simulated viewers, each scoring all twelve beats. All 179 reacted positively. Not one detractor. Per-viewer disagreement was less than half that of either comparator film in the same run — engagement SD 5.53, against 12.29 and 9.32. The panel converges, which is unusual.
Cohort engagement landed at 79.1, against 42.5 and 32.2 for the comparators. On this instrument that is a wide gap, comfortably above the point where two films become distinguishable at all.
The segment write-ups say the same thing in words. A simulated viewer in the 45–54 casual-consumer segment: "This is exactly the kind of smart, warm storytelling that makes me trust a brand. The attention to detail — both in what Wimbledon actually does and how this is made — feels genuine, not performative." Every one of the six segments came back fully positive, with no neutral and no negative share.
So the simulated verdict is unambiguous. If you watch this film, you like it.
The four seconds that decided it
Then we looked at where it peaks, and where it doesn't.
The weakest four seconds of the film are its first four. 65 out of 100 — the lowest score anywhere in 143 seconds, against a peak of 88. Those four seconds are brand supers over a wide shot of children walking onto the court. Nothing has happened yet. The first cut lands at 4.48 seconds.
On a skippable in-stream placement, the skip control appears at five seconds.
65 / 100
88 / 100Everything most viewers saw
A slow establishing wide, brand supers, and a first cut at 4.48 seconds. One frame later the viewer has a button.
Real channel data
The channel agrees
Simulated response and channel reporting are two different kinds of evidence, so we keep them in separate sentences. Here is the second kind, straight from the platform.
During the paid flight, 98,998 views produced 8 likes. In the month after it ended, 822 people found the film on their own and left 19 likes. Same film, same edit, same first four seconds.
A like rate is not a measure of how good a film is. It is a measure of whether someone was still watching and willing to act. On paid delivery, almost nobody was.
Where the film earns its attention
Twelve beats, each scored by all 179 viewers. Read as a curve rather than a total, it tells you where the edit is working and where it is spending time it hasn't earned.
Three things stand out.
The opening is 23 points below the peak, and the film takes 19 seconds to close that gap. Every other beat scores 72 or higher.
The small moments outperform the set pieces. A girl with binoculars at 59s scores 82. Children playing trumpets at 75s scores 84. Both beat the ten-second teamwork montage that sits between them.
Processing effort climbs through the second half as the interviews get more technical — 24.9 on the binoculars beat, 42.4 by the time Rufus the hawk arrives. That is our brain load index. Half of it is a function of how much is being said, so treat it as effort, not as brain activity.
| Window | Beat | Engagement | Attention | Resonance | Brain load |
|---|---|---|---|---|---|
| 0–4s | Brand supers, children cross the court · weakest | 65 | 72 | 63.9 | 36.2 |
| 4–11s | Lily and Oliver introduce themselves | 75 | 80 | 71.1 | 30.6 |
| 11–19s | Montage of tiny professionals inspecting | 80 | 85 | 76.5 | 33.4 |
| 19–26s | VIPs, VVIPs, VVVIPs · peak | 88 | 90 | 80.0 | 26.2 |
| 26–36s | Teamwork, stadium, groundskeeper | 80 | 82 | 80.6 | 38.5 |
| 36–44s | Tiny Chiefs announce the interviews | 78 | 80 | 77.1 | 30.0 |
| 44–59s | Afternoon tea, the magic ingredient | 72 | 75 | 74.5 | 37.8 |
| 59–62s | Binoculars, the views | 82 | 85 | 81.0 | 24.9 |
| 62–75s | Ball boys and girls, uniforms | 79 | 81 | 78.0 | 40.8 |
| 75–84s | Trumpets on the court | 84 | 87 | 82.6 | 24.9 |
| 84–101s | Grass at eight millimetres | 85 | 88 | 82.7 | 41.3 |
| 101–122s | Rufus the hawk | 80 | 83 | 80.0 | 42.4 |
Median scores, 0–100, from 179 simulated viewers. Simulated response, not measured human viewing. The final 21 seconds of the 143-second film are outside the twelve-beat cap and were never scored.
Targeting was not the lever
The first question a media team asks is whether a different audience would have behaved differently. We tested that on a second run: seven declared target audiences belonging to a premium menswear brand, 468 simulated viewers, the same film.
All seven landed between 76.3 and 78.4. That is a spread of 2.0 points, against the 1.78 points this instrument needs before it can call two groups genuinely different. The audience definitions are distinct — pairwise Jaccard overlap of 0.05 to 0.08. Their response to this film is not.
Read that as good news about the creative and bad news about the media plan. No audience was hiding in the targeting mix that would have sat through the first four seconds. The lever was in the edit.
Two limits worth stating out loud
Both of these apply to the numbers above. Neither is a footnote.
The last 21 seconds were never scored
The engine caps at twelve beats, which covered 0–122s of a 143-second film. The ending, the approval payoff and the brand sign-off sit outside the measurement. That is the shaded band on the chart at the top, and we are not cropping it out.
The level is uncalibrated against real people
This film scores 79.1 where the comparators score 42.5 and 32.2 — a real gap on this instrument. But the one pre-registered test of this metric against measured human viewing returned a Spearman rho of −0.142, against a pass gate of +0.388. A high score means the simulated panel liked it. It is not yet evidence that real audiences will.
And one thing we do not produce at all: brand lift or brand recall. Both are between-arms measurements — you need an exposed group and a held-out control, and a simulated panel has neither. Six fields come out of a run per viewer per beat: engagement, attention, resonance, intent, cognitive load, brain load. No recall, no lift. Anyone quoting you a simulated lift figure is quoting you an invention.
What we would change
Three edits, in the order we would make them.
Put a joke in the first four seconds
The VVVIPs exchange works as a cold open. So does any of the tiny-professionals montage, which scores 80 at 11s. The establishing wide and the brand supers can follow once the viewer has a reason to stay.
Cut a separate edit for skippable in-stream
Keep all 143 seconds for owned and social, where people arrive by choice — the 2.31% like rate says the long film works for them. The paid placement needs a version built around the five seconds it is actually given.
Stop refining the targeting on this asset
Seven declared audiences, 2.0 points apart, on an instrument that resolves 1.78. Whatever budget is going into audience refinement here is buying a difference nobody can measure. Spend it on the opening.
Great creative, delivered to people who never got to the good bit. The distance between the two was about four seconds.
Method and provenance
Key takeaways
→ All 179 simulated viewers reacted positively · peak 88/100 at 19–26s · weakest beat 65/100 at 0–4s
→ The skip control appears at 5s. The film's payoff arrives at 19s
→ Channel data: 98,998 paid views produced 8 likes (0.0081%); 822 self-selected viewers produced 19 (2.31%)
→ Seven declared audiences fell within 2.0 points on an instrument that resolves 1.78 — targeting was not the variable
→ Whole teardown: $0.65 of engine time, one afternoon
Upload a cut and get a second-by-second read from a simulated audience before the media money is committed. Visit north-ai.com.
Tags: #CaseStudy #CreativeTesting #PreLaunchTesting #AttentionIntelligence #VideoAdvertising
Run this in the simulator
Drop in a video, image, or copy. Get attention, engagement, and emotional response across 7 platform models in minutes.