Methodology · v0.3

How AdTax Radar measures ad experience

We would rather be precise than impressive. This page explains every step from Google’s data to a ranked prospect.

1. Source data

All metrics come from the Chrome UX Report (CrUX), which aggregates real-user data from opted-in Chrome users. In September 2026 Google added four experimental ad metrics (announcement, measurement methodology):

• Ad Count: average number of ad frames visible in the viewport (sampled once per second).
• Ad Density: average share of the viewport covered by ads (overlapping ads counted once).
• Ad Weight: CPU: cumulative CPU time used by ad frames and workers during the visit.
• Ad Weight: Network: cumulative kilobytes transferred by ads during the visit.

Values are 75th percentiles (p75) at origin level, with phone and desktop reported separately. Google sets no “good” or “poor” thresholds for these metrics, and neither do we. CrUX includes only sites with sufficient traffic and, for ad metrics, only pages with ads and at least one authorized seller in ads.txt.

2. Peer comparison

Each site is compared with other publishers in the same category on the same device, using mid-rank percentiles, so tied values (common for Ad Count, which is reported in coarse steps) get a middle rank, not the top. If a category has fewer than 15 sites with data for a metric, we compare against all publishers on that device and say so. Phone and desktop are never blended: ad CPU and network weight add up over a whole visit, so desktop sessions are typically heavier.

3. Change detection

CrUX reports rolling 28-day windows, and consecutive weekly windows share 21 days. We only compare windows that share no days: the latest window against the one four weeks earlier. Ad Count moves under 30% are labelled noise. A change above 100% in one window is flagged for verification and down-weighted until it persists.

4. Opportunity score

Score = 70% heaviness (weighted peer percentiles: density 30%, CPU 30%, count 20%, network 20%, renormalised over the metrics available) + 30% trend (mean worsening across metrics, capped at +50%, halved unless at least two metrics worsened). Sites without trend data are ranked in a separate “level only” tier and never scored as if nothing had changed. Missing data is always shown as missing, never as zero. Every score ships with its evidence: values, percentiles, peer counts and collection periods.

5. What a score is not

A high score means a site’s ad experience is measurably heavier, or getting heavier, than its peers in real Chrome sessions. It is not a measure of revenue, yield, losses or the quality of anyone’s monetization strategy. The metrics are experimental, and Google may change their definitions. When it does, we break the trend series rather than compare across the change.

Attribution & licensing

Contains data from the Chrome UX Report by Google, licensed under CC BY 4.0. We modify it by computing percentiles, changes and scores. AdTax Radar is not affiliated with or endorsed by Google.