RAW Chemistry — 500 creatives, blind-scored
500 typeset creatives · judges A, B · ranked by blind score. 0 agreed by both judges as launch-eligible leads. Anchor drift across batches: sd 0.30–0.58. Tap any card for the full scorecard, both judges' transcriptions and the gate results.
Method: each creative was renamed to a meaningless ID and shuffled, so no judge saw a filename, version, batch, date or any prior score — version number correlates with recency, which the rubric names as a bias to exclude. Judges viewed the phone render first and locked a first-second read before seeing the larger view. They submitted 13 raw fields and 11 hard gates; the canonical calculator owned every weight and cap. Three competitor ads were hidden in every batch under ordinary IDs to measure judge drift. CTA contrast and winner-pattern match are separate advisory columns, not inputs to the score: the account's own best-selling creative measures 1.05:1 on CTA contrast. These are pre-spend predictions, not measured CTR, CVR or ROAS.
×