This field record asks: Do Meta’s free creative diagnostics (quality, engagement, and conversion-rate rankings) identify which newly introduced ads will spend and convert cheaper than their ad set? It is part of a seven-study Darwin IQ Labs series on early signals for newly introduced Meta creatives (Nos. 113–119).

Across conversion-optimized Meta ad sets from clients sampled of various industries (59 accounts; 484 newly introduced ads in a ~35-day window), Meta’s conversion-rate ranking correlates only weakly with CPA (within-ad-set Spearman -0.16, 95% CI -0.30 to -0.01). Quality and engagement rankings show no usable link. Most new ads never receive an early ranking.

Sample

Dataset EMR-2026-10A covers 59 Meta accounts and 179 conversion-optimized ad sets (purchase, lead, or registration events) observed over Meta’s ~35-day ranking retention window. Clients sampled of various industries. Unit of analysis: ads and newly introduced ads.

Method

Three Meta diagnostics are observed at the ad-day level: quality_ranking, engagement_rate_ranking, and conversion_rate_ranking. Cross-sectionally, each ad’s modal ranking is compared with its CPA relative to its ad set. Predictively, rankings in an ad’s first three live days are tested for lift on later spend retention and CPA. Confidence intervals use ad-set bootstrap resampling. Observational field contrast — not a randomized experiment.

Findings

Weak CPA link. Leaning no. Across 398 established ads, conversion ranking barely tracks CPA (within-ad-set correlation -0.16, 95% CI -0.30 to -0.01). For new ads, only 69 of 484 got a ranking in their first 3 days and 30 had enough conversions to judge, too few to test prediction.

Conversion ranking barely tracks CPA; quality and engagement do not

Within-ad-set Spearman correlation with CPA (negative would mean better ranking → lower CPA)

Conversion ranking-0.16Quality ranking0.01Engagement ranking0.06

95% CI for conversion ranking: −0.30 to −0.01. Values near 0 mean no relationship.

Source: Darwin IQ Labs, dataset EMR-2026-10A · Clients sampled of various industries

Rankings are available mainly for single-image and video ads in conversion-optimized ad sets; flexible-format image and catalog ads rarely return a non-UNKNOWN ranking.

Early rankings that do appear are stable later (same side of average ~96% of the time), but stability without predictive power is not useful for a screen.

What this record does not claim

  • No account, brand, or creative identifiers are released with EMR-2026-10A.
  • Availability of rankings is format-dependent; results do not generalize to flexible-format or catalog ads.
  • The comparison is observational over a short seasonal window.