SplitroomRun yours
Sample report

A real report, annotated.

Kids & Baby · USA · Instagram · 1000panelists. This is an actual simulation output. Nothing below was edited. The notes on the right explain what you're looking at.

How to read this report ↓
  1. The verdict. Winner, margin, confidence band. A 68/32 split is a lean, not a landslide. The report says which, so you never over-trust it.
  2. The first 3 seconds. Before anyone reasons about your ad, they perceive it. A vision pass judges what a scroller grasps in three seconds: what the eye lands on, whether the brand registers, which creative stops the thumb. When this disagrees with the panel's considered verdict, that gap is the finding.
  3. It grades itself. Three scores before a single insight: how much the vision pass saw, how varied the panel's reasoning was, and how much this verdict deserves your trust. If the creatives were too close to call, this is where the report would say so.
  4. The call. One prescriptive sentence, plus who to target with the loser. The alternative strategy isn't a consolation prize; it's a second campaign found in the same run.
  5. Who voted. The panel itself, before you trust its verdict: named personas with real counts, each one's vote, and the age/income/occupation spread. No black-box panel: you can audit who was in the room.
  6. Why it won. Every panelist names a primary reason and up to two more. Distributions should look like this: spread across dimensions. If one bar owned the whole room, we'd flag the run, not celebrate it.
  7. The scorecard. Both ads, scored on identical dimensions. The engine can't credit one ad for a badge it didn't check on the other. That symmetry is the whole architecture.
  8. Watch out. Segments that voted united but split underneath: the kind of thing that bites two weeks into a campaign.
  9. Audience re-cuts. The same hundred votes, re-cut by income, intent, and familiarity. Wide spread = targeting signal.
  10. Meet the segments. Click any row. Real profiles, their own words, their vote.
  11. The fixes. Three ranked changes for the losing creative, each tied to a measured gap, not vibes.
  12. The debate. A moderated focus-group transcript where panelists disagree on the record. The disagreement is the insight.
  13. Signals. Attention funnel, emotional response, pushback. The texture behind the number.
Recommendedmedium confidenceLibrary-informed · 5 prior tests
Winner

Creative B

Directional lean
91%panel preference·medium confidence

Every segment agreed. All 7 buying styles on this panel picked Creative B — unanimous verdicts are rare, and worth more than the margin suggests.

See the segment splits →
Creative A · 9%
Winner
Creative B · 91%

The first 3 seconds

judged at feed size, before any reasoning
Scroll-stop: Creative B

Both land clearly, but B's downward maternal gaze and warmer, softer color palette (cream and blush tones) create a more magnetic emotional pull on first scroll than A's more clinical layout and cooler blue accents; the intimacy reads faster.

A3s graspclearbrandinstant2 focal points

First read: the mother's face and baby in her arms — the emotional anchor that draws the eye before the headline.. Headline 'Gentle enough for day one' + mother-baby embrace + product bottle land together instantly; the offer (gentle baby care) is unmissable.

Little Bloom logo and name sit top-left in navy, large and legible; reinforced bottom-left and on product label.

B3s graspclearbrandinstant2 focal points

First read: the mother's downward gaze toward the baby's face — a tender, intimate gesture that pulls focus deeper into the embrace before the headline registers.. Headline 'Every cuddle deserves the gentlest care' + intimate mother-baby moment + product bottle are cohesive; the gentle-care promise reads instantly.

Little Bloom logo and name sit top-left in navy, equally large and legible; reinforced on product label and tagline.

00Verdict confidence
Creative similarity
50%
50% divergence across scored dimensions
Expected spread
moderate
Real audiences here typically split ~60-40 to 75-25
Observation
91%
high
How much Vision caught on first pass
Reasoning
48%
low
How varied the panel's reasoning was
Verdict
63%
medium
Directional lean · B 91%
Spread sanity check

The Alignment engine scored every dimension identically for both creatives — usually a sign the scoring dimensions didn't fit these ads (check the category). The similarity reading is unreliable and was excluded from confidence.

Ship Creative B

Launch Creative B. Its ingredient-forward trust proof reads faster, and the warmer emotional tone lands with parents.

The evidence

Every finding below expands. The headers alone tell the story — open the ones you want to interrogate.

These are simulated audience reactions from Splitroom's synthetic panel, not campaign performance predictions. Within this panel, Creative B was preferred by 91% for the reasons above.