This field record asks: If landing pages are stable, do cheaper clicks (lower CPC) or higher CTR predict lower CPA across ads in the same conversion-optimized 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 Meta ads in conversion-optimized ad sets from clients sampled of various industries (62 accounts; 237 ad sets), cheaper clicks convert worse (median correlation 0.82 between high CPC and high CVR). CPC explains roughly none of the CPA spread among prospecting ads (−9%, 95% CI −20% to 5%). First-week CPC lift on later CPA is 0.96×; first-week CPA (3+ conversions) lifts 1.21×.

Sample

Dataset EMR-2026-10D: 16-week weekly ad panel, 62 accounts, 237 ad sets. Clients sampled of various industries.

Method

Within each ad set, click-to-purchase dispersion is compared with binomial noise. Log CPA is decomposed into log CPC and log(clicks/conversion). Predictive checks use first-week CPC/CTR/CPA relative to the ad set against the next three weeks’ outcomes. Observational.

Findings

CPC is not a CPA stand-in. No. Cheaper clicks come with lower click-to-purchase rates (median correlation 0.82 between high CPC and high CVR), and the two nearly cancel. In prospecting ad sets, CPC explains roughly none of the CPA differences between ads (estimate -9%, 95% CI -20% to 5%). A cheap first-week CPC picks next-3-week CPA winners 0.96x as often as random; first-week CPA (3+ conversions) does better at 1.21x.

First-week CPC does not pick later CPA winners; early CPA does a little

Lift on next-3-week CPA (performs) from first-week signals

1.00x barLow first-week CPC0.96xHigh first-week CTR0.94xLow first-week CPA (3+ conv)1.21x

Source: Darwin IQ Labs, dataset EMR-2026-10D

Click-to-purchase rates vary far more than chance (median dispersion ~6.7×).

What this record does not claim

  • No account identifiers are released with EMR-2026-10D.
  • Ads reaching 20+ conversions are already favored by delivery.
  • Meta chooses each ad’s audience; CVR differences mix creative and audience effects.