This field record asks: How many impressions does a newly introduced Meta ad need before early spend share, CTR, or CPC stops rearranging relative to its 21-day outcome? It is part of a seven-study Darwin IQ Labs series on early signals for newly introduced Meta creatives (Nos. 113–119).
Sample
Dataset EMR-2026-10C: 30 accounts, 58 ad sets, 279 new ads with a full 21-day window. Clients sampled of various industries.
Method
Signals (spend share, CTR, CPC, CPM, conversion ranking when present) are measured at cumulative impression cuts of 500, 1k, 2k, 5k, and 10k. Lift compares the top third at each cut against all new ads on day-14–20 spend retention and 21-day CPA. “Useful from” is the earliest cut whose lift CI exceeds 1. Observational — not a lab randomization.
Findings
Spend share is readable early. Spend share already predicts which new ads keep spending at 500 impressions (about $10 of spend), with about 2.0x lift. CTR and CPC need about 2,000 impressions and 1,000 impressions and are weaker. No signal predicts which new ads will beat their ad set's CPA at any cut point.
Spend share predicts later delivery from ~500 impressions; CTR/CPC need more
Lift on still holding spend in days 14–20 for ads in the top third at each cut
Exact lifts vary by cut; spend-share CIs clear 1 from 500–1,000 impressions.
Source: Darwin IQ Labs, dataset EMR-2026-10C · 1.0× = random
At median CPM in these ad sets, 500 impressions cost roughly $10 and 1,000 roughly $20.
What this record does not claim
- No creative files are released with EMR-2026-10C.
- Daily data makes low cuts approximate for ads that explode on day 0.
- Requiring 10k+ impressions over 21 days drops ads Meta starved completely.