Data identity & clean rooms

Ten companies, 154 claims, 46 of them outcome claims: 14 implied, 32 stated metric, 0 guaranteed, 0 none.

What outcome means here is a named client's campaign or operating result attributed to a better-resolved customer record, with the match or recognition rate standing in for the result wherever no downstream figure is published.

Evidence on those 46 claims is 21 named clients, 11 bare numbers, 4 described methods, 2 third-party references and 8 with none; the standard shape is a client name beside a percentage, which records that the result occurred and was credited to the vendor's data.

Both third-party references are Forrester Total Economic Impact studies, one disclosed on the page as commissioned by the vendor liveramp.com and one quoted mid-clause with no figures shown dnb.com; a scan of all captured pages returns no independent accuracy audit or accreditation.

No outcome claim is guarantee-coded, no page carries make-good, refund or credit language, and the three pages that use a guarantee word attach it to governance, ID stability and clean-room neutrality rather than to a result.

Of 36 pillars, 28 carry evidence somewhere on the site, 12 of those on a page other than the front door, and 7 carry none.

Figures measured by the vendor on a file the vendor selected travel to a second buyer as an existence claim only; publishing the file, the window and the buyer's category is what would move them.

Open ground: an accuracy figure tested by a party the vendor does not pay, with a commercial term attached to missing it; 3 of 10 sites publish any pricing page, and boundary is four to five front-door pages per company in one capture week, with Epsilon's home script-rendered, Databricks' customer grid unrendered and TransUnion's method behind a gated PDF.