// 01, The ad

The opening, on the record.

A Nike YouTube spot leading with a problem-solution frame. Watch the first three seconds, the text is doing the work the picture should be doing.

Nike · YouTube · US

Click to play the full ad.

10.1s · YouTube / brand channel
benchmark_only = true · scored 2026-05-18

02 · Archived result

Directional creative evidence.

67
Archived raw composite · not outcome confidence

The archived runner returned SHARPEN. This is a historical rule label, not a current decision recommendation. No matched-cohort sample or repeated-run estimate is attached to this example.

Meta has no held-out validation. A platform match alone also does not establish coverage for a market, category or format. Keep these examples directional unless the specific cohort evidence supports a stronger statement.

// 03, The five KPIs

Where the score comes from.

Archived raw KPI outputs below. Labels describe diagnostic ranges only; they are not measured retention, recall, persuasion or emotional intensity.

// KPI breakdown, Nike YouTube US

Beat the Skip
25%
Beat the Skip, Feet on pavement, Nike swoosh on socks AND shoes AND on-screen text. brand_in_first_3s = yes, three times over.0.5s
75
Archived model diagnostic. Inspect the linked frame and the full sequence; no viewer response was measured.
Get Noticed
20%
Get Noticed, 'NIKE INVINCIBLE 3' caption on the opening frame, brand AND product lock together before the second beat.2.0s
70
Archived model diagnostic. Inspect the linked frame and the full sequence; no viewer response was measured.
Brand Impact
20%
Brand Impact, Runner in red Nike outfit, NYC streets, the swoosh in the corner stays present.4.0s
70
Archived model diagnostic. Inspect the linked frame and the full sequence; no viewer response was measured.
Sell Proposition
20%
Sell Proposition, Runner mid-stride, NYC architecture, performance proof through location and gait.MID
65
Archived model diagnostic. Inspect the linked frame and the full sequence; no viewer response was measured.
Build Brand
15%
Build Brand, Wide shot of the runner, silhouette + Nike swoosh in frame, codes carrying memorability.75%
60
Archived model diagnostic. Inspect the linked frame and the full sequence; no viewer response was measured.
// 04, Creative classification

What the pipeline tagged.

Hook type
Problem-Solution
Opening motion
Dynamic
Brand in first 3s
Yes
Attention risk
Text dominance
Value prop type
Feature-Led
CTA timing
None
Subcategory
Running
Matrix label
missed_opportunity
05 · Interpret the run

Read the model output.

The archived runner tagged the hook as Problem-Solution, opening motion as Dynamic and value proposition as Feature-Led. Check these labels against the full creative. Review whether on-screen text is readable during normal playback.

06 · Proposed revision

An edit to test.

Nike ad closing frameEND

Compare a reduced-text version with the original while retaining the demonstration. Keep the original, record the change and agree the outcome measure before testing. There is no measured score uplift or campaign result for this proposed revision.

07 · Coverage and stability

Coverage limits.

This archive identifies the platform and country, but does not attach a matched-cohort sample, a percentile calculation or a repeat-run stability estimate. We therefore do not rank this ad in a category quintile. Validation of the separate platform-native attribute model does not validate these five KPI outputs.

08 · Decision and follow-through

Record the human choice.

Use the example to frame a review question. A human reviewer should record the brief, chosen revision, proposed test and actual result. No client acceptance or in-market outcome is claimed for this case. See the methodology and the brand-cue review guide.

Archived run evidence. Scored 18 May 2026. Raw scores and classifications are preserved from the archived scoring output. Prompt hash, temperature, seed, matched-cohort coverage and repeat-run stability were not recorded in that archive. All interpretations here are directional. These are real ad assets with model-generated diagnostics, not synthetic consumer measurements. No sales, ROAS, conversion, retention, emotional-strength or brand-recall result is established. Current methodology and separate model-validation evidence.