built from JUDGE-FRAMEWORK §5 · 95 competitor creatives (39 Goda + 56 Cupids, 7h–535d) · 2 model judges: gpt-5.1, gpt-4.1
The framework said this judge had never been built. §6: "The judge is
not built. This is the specification." It is now — 7 men's-market dimensions, the §5c
calibration anchors, and the miscalibration gate that must pass before any score is trusted.
CALIBRATION FAILED — 5 of 6 anchors out of band, so the absolute scores below
are NOT trustworthy. The framework's own rule (§5c): "if it scores CUP-13/14/15 below RAW's
existing pool, it is wrong." It scored that 68-day LIVE winner at 7.0 against an expected
8.5+. The judge is harsh across the board — the exact failure §5c exists to catch.
Root cause found, and it's a defect in the spec, not the code. §1 states Goda sells to women
and "the devices do not transfer" to RAW's men's market. §5c then anchors two Goda
creatives at very high. A judge built faithfully from both cannot pass — it correctly
discounts Goda devices per §1, then fails the Goda anchors from §5c. It scored the Goda 508d anchor
4.75 against an expected 8.5+. The anchor set must be rebuilt from Cupids only.
Calibration test
Anchor
Expected band
Range
Scored
CUP-13 FREE JUST TODAY! $120 -> $0
very_high
8.5–10.0
7.0
FAIL
GODA 508d LIVE: Do Not Try This IF you enjoy being the only one
very_high
8.5–10.0
4.75
FAIL
CUP-01 MAKE HER KNEES WEAK (195d, longest Cupids static)
high
7.5–9.0
6.5
FAIL
CUP-03 LAST CHANCE - BUY 2 GET 1 FREE (176d)
high
7.5–9.0
6.65
FAIL
CUP-18 "I'm wet Daddy." - Your Wife (56d live)
good
6.5–8.0
6.25
PASS
GODA 432d square product-hero, no proof, dead
medium
5.0–6.8
1.35
FAIL
Take-5 copy, scored (indicative only — calibration failed)
Overall
Concept
Headline
Offer as hook
Tension
Outcome
Register
Verdict
6.55
H-SEQUIN
DON'T WEAR THIS IF YOU WANT TO BE LEFT ALONE.
0
8.5
7
7
Excellent reverse-warning hook for male buyers; add a price hit or ‘50% OFF TODAY’ bar and stars to make it a
6.3
LOOKUP
WARNING: SHE WON'T LEAVE YOU ALONE.
0
7.5
7.5
6.5
Strong warning hook and clingy outcome; with a hard discount/‘free bottle’ callout and rating strip it’s a sol
6.15
CARBON-SEQUIN
F*CK YOUR COLOGNE. WEAR THIS.
0
7.5
7
6
Punchy attack on generic colognes with clear sexual frame; bolt a price hook and proof bar to make it viable.
6.15
CUPID-R4B
SHE WANTS YOU IN THIS. NOTHING ELSE.
0
6.5
8
6.5
Very male-ego flattering and concrete; strong foundation that just needs an explicit sale/discount and proof l
5.85
CUPID-R2
MAKE HER LOSE IT.
0
5.5
7.5
6.5
'Make her lose it' is clear and reusable but generic; wants an in-headline offer or number and stronger visual
5.3
IM8-R5
THE ONLY COLOGNE BUILT FOR SEDUCTION.
1
4
5.5
7
Positioning line plus decent proof bar; needs price-as-hook and a sharper, more observable sexual outcome to c
5.0
CARBON-CLUB
I WASN'T LETTING HIM LEAVE.
0
5
7.5
5
Strong sexual reaction visual, good tension; needs offer/price and proof strip to be men’s-market ready.
4.4
GODA-R4B
I COULDN'T GET OFF HIS NECK.
0
4
7
4.5
Highly visual, explicit neck fixation; good for men but still missing the aggressive offer that drives this ma
One finding survives the calibration failure, because it's structural rather
than a score level: every single one of our eight headlines scores 0–1 on "offer as hook."
In the men's-market corpus that is the dominant winning device — FREE JUST TODAY!
$120→$0 (68d live ×3), LAST CHANCE — BUY 2 GET 1 FREE (176d), 50% OFF —
SALE ENDS TONIGHT (182d). We put the offer in a small pill and gave the headline to mood.
The evidence says for men, the offer is the hook.