Methodology
How the numbers on this site are measured
Two independent sources, neither of them anybody’s sales copy: each company’s own sellers.json, and the ads.txt files of real publisher sites. Every number is tied to a file you can open yourself.
Which companies are measured?
Every company comes from Google’s Certified Publishing Partner directory: 203 entries, 199 with a usable domain, fetched 26 August 2026. If a company is listed there, it is marked GCPP. Six of them are Premier partners.
What does sellers.json show?
Each company’s sellers.json was fetched at https://[domain]/sellers.json; 131 parsed and 68 returned an error. Publisher domains are the unique domains in that file, counting PUBLISHER and BOTH entries. The ratio is distinct seller IDs divided by publisher domains.
| Ratio | What it means for you |
|---|---|
| 0.7 and above | Each publisher tends to get their own account. Your impressions, eCPM and fill are yours to read. |
| 0.2 to 0.7 | Mixed. Some publishers get their own ID, some share one. |
| below 0.2 | IDs are shared across many publishers. You see a share of a pooled number, not your own. |
Where a company’s sellers.json returned an error or wasn’t at the standard path, every measured column is a dash. No estimates, no filling in gaps.
What does the Indian ads.txt scan show?
I read the ads.txt file of 123 Indian publisher sites — national news, regional language, tech, sport, entertainment — and counted which partners appear with a DIRECT line. DIRECT means the publisher has a direct relationship; a RESELLER line only means inventory reaches them through somebody else.
What else was scanned?
A worldwide ads.txt scan: 2,200 publisher domains seeded across 7 regions, of which 1,165 returned a real text/plain ads.txt. Saturated exchanges were removed — google, pubmatic, rubicon, appnexus, openx, criteo, taboola, outbrain, indexexchange, sovrn, triplelift, sharethrough and 37 others. A DIRECT line means the publisher has an account with that company, not that anyone manages its inventory.
What are the limitations?
- Worldwide seeds are biased: they were sampled from publisher domains already listed in some parent’s sellers.json, so they over-represent publishers who already have a partner.
- 68 certified companies have no measured numbers. A blocked file is not a small company.
- Confidential sellers are excluded, so some publisher counts are floors, not totals.
- Geography is a TLD proxy, not a company headquarters and not where their revenue comes from.
- Absence means not found in this sample, not that they never work with anyone in India. 123 sites is a sample, not a census.
When were the numbers checked?
All files were fetched on 26 August 2026. The ten partner reviews were re-checked against each company’s current sellers.json on 15 September 2026, and each review shows what changed.
Why this way?
I compete with these companies. That’s the reason every number here is tied to a file you can open yourself rather than to my opinion. The full data is in the Google MCM (Multiple Customer Management) partners directory.