This field record estimates how often Meta video impressions clear the platform’s three-second video-view threshold, and whether ads that hold attention past three seconds also show higher link click-through. The short name used in the repository directory is the 3-second video view drop-off study.

Across n=1,152 Meta video ads from clients sampled of various industries in July 2026 auctions, 31.9% of impressions register a three-second video view—so 68.1% drop off before three seconds. Ads in the highest quartile of three-second view rate show 56% higher impression-weighted link CTR than the lowest quartile (Pearson r = 0.35). After removing each advertiser’s mean CPM, the association between three-second rate and CPM is weak (within-account r = 0.18).

Sample

The asset set is dataset EMR-2026-08: 1,152 Meta video ads with measurable spend and Meta video_view (three-second) actions in July 2026, drawn from the same random sample of advertiser accounts as EMR-2026-07B (accounts that each spent at least $100,000 in June 2026). Clients sampled of various industries. Ads required at least $50 spend and 1,000 impressions in the window. Creative assets were not downloaded; format was classified from creative metadata only.

Method

The primary outcome is the impression-weighted three-second view rate: Meta three-second video views ÷ impressions. Drop-off before three seconds is one minus that rate. Secondary analysis correlates ad-level three-second view rate with link CTR (inline link clicks ÷ impressions). Tertiary analysis correlates three-second view rate with CPM, reporting both a pooled correlation and a within-account demeaned correlation so advertiser-level auction prices do not dominate the estimate.

Findings

In this July 2026 Meta window, fewer than one in three video impressions reach the three-second view threshold on an impression-weighted basis (31.9%). The unweighted ad-level median is lower (22.9%), indicating a long right tail of high-hold assets.

Most video ads lose most viewers before three seconds

Number of video ads by 3-second view rate, Meta auctions, July 2026

050100150200 adsMedian ad: 22.9%All impressions: 31.9%A long tail of high-hold ads0%25%50%75%100%Share of an ad's impressions that reached a 3-second view

Note: 3-second view rate is Meta three-second video views ÷ impressions. The 31.9% figure weights each ad by its impressions.

Source: Darwin IQ Labs, dataset EMR-2026-08 · 1,152 Meta video ads with at least $50 spend and 1,000 impressions · Clients sampled of various industries

Link CTR rises with three-second hold: quartile 1 CTR is 1.06% and quartile 4 is 1.65% (index 156 vs 100). The Pearson correlation between three-second view rate and link CTR is 0.35 pooled and 0.33 after within-account demeaning—a moderate positive association, not a tight deterministic relationship.

Ads that hold attention tend to earn more clicks

Link click-through rate vs. 3-second view rate; each dot is one video ad

0%1%2%3%4%5%6% link CTRQ1 · 1.06%Q2 · 1.00%Q3 · 1.49%Q4 · 1.65%Correlation: r = 0.35Blue: quartile averages, weighted by impressions0%25%50%75%100%3-second view rate

Note: 56 ads with link CTR above 6% are drawn at the top edge. Q4 link CTR is 56% higher than Q1 (index 156).

Source: Darwin IQ Labs, dataset EMR-2026-08 · 1,152 Meta video ads, July 2026

Pooled CPM and three-second rate are essentially uncorrelated (r = 0.02). Within-account demeaning raises that to r = 0.18. Advertiser price levels therefore explain most of the naive CPM pattern; residual within-account association remains weak.

Hold time tracks clicks, not auction prices

Correlation between an ad's 3-second view rate and its link CTR or CPM

NegligibleWeakModerateLink CTRAll ads pooled0.35Link CTRWithin each advertiser0.33CPMAll ads pooled0.02CPMWithin each advertiser0.180.00.10.20.30.40.5Correlation with 3-second view rate (Pearson r)

Source: Darwin IQ Labs, dataset EMR-2026-08 · “Within each advertiser” removes each account's mean before correlating, so price differences between advertisers do not drive the estimate

What this record does not claim

  • No creative-level file, account list, or brand identifiers are released with EMR-2026-08.
  • No confidence interval or causal estimate of hold on CTR after controlling for objective, placement, or audience is part of this release.
  • The comparison is an observational field contrast in live Meta auctions, reported as observed. Three-second views follow Meta’s video_view definition and are not laboratory attention measures.