Evidence
Cohorts, runs and scripts
Nothing on this site is a number we were told. It is a call we made, with the response kept.
Cohort runs
| Run | Cohort | Advertisers | Calls | Valid | Est. cost | Raw responses |
|---|---|---|---|---|---|---|
| 2026-09-15 | v1 | 10 | 100 | 100 | $1.58 | One JSON file per call at /evidence/ads-runs/2026-09-15/<platform>-<provider>-<aNN>.json, e.g. meta-sc_company-a01.json; tables as CSV under /evidence/ads-runs/2026-09-15/tables/. |
Re-run the cohort
treg auth login
scripts/run-ads-cohort.py evidence/fixtures/ads-cohort-v1.json evidence/ads-runs/$(date +%F)
scripts/score-ads-cohort.py evidence/fixtures/ads-cohort-v1.json evidence/ads-runs/$(date +%F)Download cohort v1 · runner · scorer
What is in the raw files
Each file holds the provider's full response plus a run header (platform, provider, advertiser, wall-clock latency, timestamp, exit code). Ad-library records are public data published by the platforms about advertisers; creative URLs point at the platforms' own CDNs and may expire. No credentials and no personal data are present.