/* global window */ /* Report content object — consumed by blog-report-tpl.jsx via window.__REPORT__. Fully anonymized: a national food & beverage brand's pre-roll creative test. */ window.__REPORT__ = { slug: "food-ad-emotion-test", title: `Sample Report: Two ad creatives, 244 faces — what emotion data said before launch`, lane: [`CONSUMER / FOOD & BEVERAGE`, `FIELDED FEB 2025`], heroTags: [ { label: `SOURCE-LINKED` }, { label: `ANONYMIZED` }, { label: `n = 244` }, { label: `VIDEO EMOTION` }, ], heroMeta: [ { dt: `METHOD`, dd: `Webcam emotion capture, frame-by-frame` }, { dt: `SAMPLE`, dd: `244 usable (1,334 eligible)` }, { dt: `STIMULI`, dd: `2 in-house cuts + competitor benchmarks` }, { dt: `AGE SPLIT`, dd: `13–17 (44.7%) · 18–24 (55.3%)` }, { dt: `GENDER`, dd: `68% female · 27% male · 5% other` }, { dt: `SHIPPED`, dd: `Brief to verdict, pre-launch` }, ], summaryPre: `Across 244 usable reaction videos, the predominant expression was flat and slightly negative — exactly what a distracted viewer looks like, and a poor verdict on the creative. So we measured change off each viewer's own baseline, not raw intensity. Read that way, smiling is the channel that tracks ad and brand effect`, summaryMid: `, and at scale a positive shift is recoverable from webcam responses well enough to predict ad liking`, summaryPost: `The one reaction worth acting on was a Surprise spike — invisible in the average, isolated to the 13–17 segment, on a single second of one cut. Stated preference would never have found it.`, cites: { c1: { label: `[FACIAL CODING · 2023]`, ref: `HÖFLING & ALPERS 2023 · FRONTIERS IN NEUROSCIENCE · 219 PARTICIPANTS, 64 COMMERCIALS`, body: `Automatic facial coding of smiling predicted self-reported emotion, ad likeability, brand likeability, and purchase intention — with incremental value beyond self-report alone. Evidence that the "Happy" channel, not the dominant resting expression, is the signal that tracks advertising effect.`, }, c2: { label: `[ROC AUC 0.85 · 2015]`, ref: `McDUFF ET AL. 2015 · IEEE TRANS. AFFECTIVE COMPUTING · 12,000+ RESPONSES, 1,223 PEOPLE, 170 ADS`, body: `Ad liking was recoverable from webcam facial responses at ROC AUC = 0.85, and a change in purchase intent at 0.78. The operative caveat for this report: positive expression converts best when it lands immediately before the brand appears. Timing is the signal, not the average.`, }, }, stats: [ { v: "244", l: "usable reaction videos analyzed" }, { v: "1,334", l: "eligible participants funneled" }, { v: "1", l: "cut that earned a real reaction" }, { v: "13–17", l: "the segment that surprised" }, ], findingsHeading: `Five findings — and only one of them is a reaction.`, findingsIntro: `Each finding is read off the time-aligned emotion trace, normalized against each viewer's own baseline. The resting face is a floor, not a verdict; what counts is where the face moves, in which direction, and at which second. The full segmented traces ship in the downloadable report.`, findings: [ { num: `FINDING 01`, tag: `HIGH IMPACT`, tagAccent: true, hi: true, title: `Measure change, not the dominant emotion`, verbatim: `Sad was the floor. Happy was the only channel that actually moved with the content.`, meta: [`baseline-normalized`, `Happy = signal`, `Sad = canvas`], timeline: [0.08, 0.31, 0.55, 0.79], }, { num: `FINDING 02`, tag: `CONTEXT`, title: `The predominant expression was flat`, verbatim: `Across all 244 videos the most common expression was low-grade negative, and it did not develop.`, meta: [`whole population`, `did not develop`], confidence: 0.74, confidenceLabel: `SHARE AT NEGATIVE BASELINE`, }, { num: `FINDING 03`, tag: `LOW RISK`, title: `Negative emotion never spiked`, verbatim: `Fear and Anger showed no meaningful shift for any clip — no aversion event, no recoil.`, meta: [`Fear + Anger channels`, `no harm`], confidence: 0.12, confidenceLabel: `PEAK NEGATIVE SHIFT`, }, { num: `FINDING 04`, tag: `HIGH IMPACT`, tagAccent: true, hi: true, title: `The Surprise spike — one cut, one segment, one second`, verbatim: `A sharp Surprise spike appeared only in 13–17 White & Asian viewers, on cut one, at the cup-placement moment — then a positive follow-through.`, meta: [`13–17 subgroup`, `cut one only`, `Surprise → Happy`], confidence: 0.45, confidenceLabel: `13–17 SHARE OF SAMPLE`, }, { num: `FINDING 05`, tag: `MED IMPACT`, title: `Cut two moved no one; benchmarks set the ceiling`, verbatim: `The reordered second cut drew no strong reaction in any demographic; the longer competitor spots with people and animals pulled the strongest Happy in the test.`, meta: [`cut two flat`, `benchmarks calibrate`], confidence: 0.05, confidenceLabel: `CUT-TWO PEAK SHIFT`, }, ], recommendation: { heading: `Lead recommendation surfaced from Finding 04`, title: `recommendation`, body: `Build the launch around the one moment that earned a reaction. Cut a short version that front-loads the placement-and-swirl beat before the brand mark appears — positive expression converts to intent only when it lands right before the brand. Retire the second cut. If there is budget for a follow-up, the longer story-and-character structure of the benchmarks is the proven pattern.`, tags: [`SURPRISE → HAPPY`, `YOUTH 13–17`, `FRONT-LOAD THE BEAT`, `244 VIDEOS`], confidence: 0.7, confidenceLabel: `EVIDENCE STRENGTH`, }, evidenceHeading: `Three signal traces behind the verdict.`, evidence: [ { time: `~0:06`, speaker: `SURPRISE CHANNEL · AGES 13–17`, body: `Sharp percentile spike at the cup-placement second on cut one, then a positive follow-through. Flat for every other clip and every other age group — the entire case for video over a preference question.`, tags: [`SURPRISE · 13–17`, `CUT ONE`, `FINDING 04`], }, { time: `—`, speaker: `HAPPY CHANNEL · BENCHMARK SPOTS`, body: `The strongest movement off baseline in the whole test, by a wide margin. The benchmarks are not the deliverable; they prove the measurement can detect a strong reaction when one exists.`, tags: [`HAPPY · BENCHMARK`, `CALIBRATION`, `FINDING 05`], }, { time: `—`, speaker: `SAD CHANNEL · FULL POPULATION`, body: `Highest absolute intensity in the session, near-zero change across clips. The resting face of a distracted viewer — a floor to measure movement against, not a verdict on any creative.`, tags: [`SAD · POPULATION`, `THE FLOOR`, `FINDING 02`], }, ], personasHeading: `Five segments — who reacted, and to what.`, personasIntro: `Segmented from the time-aligned emotion traces by age, gender, and ethnicity. Shares are of the 244-video sample; the line under each is the reaction that defines it.`, personas: [ { share: `~20%`, tag: `MAIN PERSONA`, name: `The 13–17 Surpriser (White & Asian)`, quote: `The only segment that flinched — Surprise into Happy at the cup-placement second on cut one, and nowhere else.`, about: `The youngest viewers, most likely to share. The single actionable reaction in the test lives here. Build the cut-down for them.`, }, { share: `~25%`, tag: `NO SIGNAL`, name: `The 13–17 Unmoved (Black & Hispanic)`, quote: `Cut one did not resonate. No strong shift either way, positive or negative.`, about: `Same age, different response. The placement gag did not land. A reminder that "youth" is not one audience on this creative.`, }, { share: `~55%`, tag: `MUTED`, name: `The 18–24 Baseline`, quote: `Flat across both in-house cuts; the benchmarks still moved them.`, about: `The larger half of the sample, and the quieter one for this creative. They prove the sensor works — they react to the strong benchmarks — they just do not react to the cuts on test.`, }, { share: `27%`, tag: `MOST RESPONSIVE`, name: `The Male Responder`, quote: `Most responsive overall, with the strongest Happy of anyone on the character-and-animal benchmark.`, about: `A smaller slice of the sample but the most expressive. Useful for calibration; not where the in-house cuts won or lost.`, }, { share: `68%`, tag: `NO DIFFERENCE`, name: `The Female Majority`, quote: `Reacted, but with no actionable creative-level difference between the cuts.`, about: `Two thirds of the sample. They engaged, yet the in-house cuts did not separate for them — which is itself a finding when a creative is meant to.`, }, ], gateTitle: `Get the full creative read — every emotion channel, all five segments, and the time-aligned trace behind each finding.`, gateDesc: `Drop your work email. We'll send the full report, including the segmented Surprise and Happy traces and the funnel from 1,334 eligible to 244 usable videos (PPTX and CSV available on request). No marketing follow-up unless you ask — we hate it too.`, gateIncluded: [ `Full PDF`, `Segmented emotion traces`, `Funnel + demographics breakdown`, `CSV channel export`, ], related: [ { href: "/blog/sample-report.html", report: true, lane: "REPORT · 13 min", date: "Oct 2025", title: "The challenger's playbook: beating the Giants in sportswear (n=259)", excerpt: "Source-linked findings, five personas, and the verbatim evidence behind each — how a challenger steals share from incumbents.", author: "NeroView Research", meta: "Consumer / Sportswear", }, { href: "/blog/synthetic-vs-human.html", lane: "INDEX · 10 min", date: "Feb 2025", title: "Synthetic respondents vs. human evidence: where each one earns its place", excerpt: "When an AI persona is a fast sketch and when it quietly lies to you — and why a real face beats a simulated opinion on a pre-roll.", author: "NeroView Research", meta: "Methodology", }, ], };