Methodology

Anonymized, aggregated account-level paid-media metrics from Synter-connected ad accounts (TimescaleDB ad_metrics warehouse).

Aggregation window: trailing 0 days per edition (2026-01-01 → 2026-03-31 for Q1 2026). Accounts with under $50 of spend in the window are excluded.

Account-weighted: each account contributes at most one observation per platform/segment/quarter cell.

We report account-weighted percentiles rather than spend- or impression-weighted aggregates, so a handful of large accounts cannot dominate the benchmark. Each account is one vote.

Medians and percentiles (p10/p25/p50/p75/p90); means are not reported.

Account-level values are winsorized to the 1th–99th percentile before summarizing, limiting the influence of extreme outliers. We publish medians and percentiles (p10/p25/p50/p75/p90); we do not report means.

95% confidence interval via account-level bootstrap (2000 resamples).

A metric is only computed for an account when its denominator clears a floor, so noisy ratios from tiny volumes never enter a cell:

Cells failing minimum-distinct-account or dominance thresholds are suppressed, never published, regardless of statistical confidence. No single-account trajectories are exposed.

• Minimum 20 distinct accounts per published cell (k-anonymity floor).
• A cell is suppressed if a single account exceeds 40% of the cell's spend.
• A cell is suppressed if the top 3 accounts exceed 70% of the cell's spend combined.

Industry/vertical segmentation is not yet published (requires an account->industry mapping); segments are objective classes derived from campaign objective. Platform-reported conversions are used; they differ from downstream pipeline.

• Q1 2026 — 2026-01-01 → 2026-03-31 · artifact 2026-q1.1