Why it won, with the receipts.
Every winning ad gets a plain-English story of what carried it. The numbers are computed in code first, and the AI is only allowed to cite them, so an invented statistic physically cannot reach the screen.
The taggings, and the insight each one produces.
Every ad is read down to its genome, 140+ elements across 21 families. Then it isolates the one element that actually carried the win. Here it is, in full, on three very different accounts.
Illustrative reads on demo brands. On your account, every number is your own real data.
Lumé · Skincare · UGC selfie
The bare-skin close-up is the part that stopped the scroll.
Hook
44 +11
Hold
31 -3
LP match
72 +8
The genome read
140+ elements · 21 families
+ 15 more families read on this ad, each with its own confidence and evidence.
Forge · Fitness · Ingredient macro
The quantified claim, 5g and zero chalk, is what stopped the scroll.
Hook
47 +12
Hold
29 -2
LP match
74 +9
The genome read
140+ elements · 21 families
+ 15 more families read on this ad, each with its own confidence and evidence.
Vault · Fintech · Review card
The specific-dollar review, the $240 line, is what stopped the scroll.
Hook
42 +8
Hold
34 +2
LP match
79 +14
The genome read
140+ elements · 21 families
+ 15 more families read on this ad, each with its own confidence and evidence.
The AI can't make up a number. It's not allowed to.
Most “why did this win” tools just ask a chatbot and hope. Win Story is built the opposite way, in two strict stages.
Code does the math first
Before any AI runs, code reads the ad's full genome against every other ad in your account and computes the evidence: the lifts, the cohorts, the isolated effects, the scores versus your own medians.
The AI can only cite it
Then the model writes the story, but it can only reference the pre-computed evidence by id. It can say what a number means. It cannot make one up.
The screen enforces it
Any claim citing evidence that doesn't exist is stripped, and every number on screen is pulled from the evidence pack, not the prose. Fabrication has no path to your eyes.
It separates what it can prove from what it just noticed.
The hard part of “why it won” is that winners co-occur. An ad had a founder hook AND a discount AND won, so which one carried it? Every receipt in a Win Story is stamped with how sure it is.
Proven
The element still wins once we hold everything else in the ad fixed. Not a lucky co-occurrence, a real driver you can scale.
Correlated
A strong pattern across a cohort of your ads that share this element, but not yet isolated from confounds. Worth leaning into, with eyes open.
Observed
A true fact about the ad with no stats attached yet, often a fresh bet the account has never tried. Flagged as an early signal, never dressed up as proof.
And it grades every ad against your own account's history, never an industry average.
Get the receipts on your own winners.
Book a 30-minute teardown. I'll run a Win Story on your best ads and show you which element actually carried the win, with the proof.