Every ID modality is a claim about who’s behind the screen. Accuracy is simply how close the claim gets to an actual person. Here is what the claims are actually worth, measured against ground truth.
Most open-internet identity spend still lives in the first two eras. Validation is the other side of the gap, bringing walled-garden certainty to the open internet and CTV.
Linkage accuracy: CIMM identity benchmarking research (commissioned by Truthset) and Truthset estimates.
An independent benchmark, commissioned by Truthset, of the identity linkages that move billions in programmatic spend. The results:
| Linkage | Measured accuracy | What that means |
|---|---|---|
| IP → postal | 13% | Roughly 1 in 8 IP-to-home linkages points at the right household. The rest of your geo-keyed targeting is noise. |
| IP → email | 16% | The bridge most CTV identity quietly rides on. Wrong 5 times out of 6. |
| Email (HEM) → postal | 32–69% | Across 15 major providers, the same linkage ranges from worse-than-a-coin-flip to decent. You don’t know which one you bought. |
| Email → postal (average) | 51% | The market average, per Truthset’s State of Data Accuracy 2026: a coin flip. |
| Provider agreement | 6.4% | How often identity providers even agree with each other on the same records. There is no consensus reality under the market’s match rates. |
Sources: CIMM identity benchmarking research (commissioned by Truthset); Truthset State of Data Accuracy 2026.
Truthset’s State of Data Accuracy 2026 is the culmination of years of research measuring the accuracy of consumer data across the industry’s largest providers.
It found error rates of up to 60% in demographic data, the attributes campaigns are targeted, priced and measured against, alongside the identity-linkage gaps above.
Compound bad identity with bad demographics and the bill arrives: the report projects that inaccurate demographic and identity data will waste $7.4 billion in programmatic CTV advertising.
And AI-driven optimization doesn’t fix bad signals. It gets more efficient at being wrong, faster, at scale and with confidence.
error rates found in common demographic attributes
projected annual waste in programmatic CTV
provider-to-provider agreement on the same records
A typical activation path re-keys your first-party file four times before an impression serves: onboarder translates PII to device IDs, a probabilistic graph models the joins, a reseller re-keys again, and the DSP/SSP trades on bid-stream IDs. Each hop multiplies the linkage error rates above, which is how a file of real customers becomes an audience that is mostly guesses, and how $1.00 of media becomes roughly $0.12 of on-target delivery.
The fix is fewer translations. One validated 1:1 PII match between your file and activation, with every linkage scored before it is used. That is the ID Max path, and roughly $0.93 of every dollar lands on the people you intended.
If your identity partner can’t tell you these numbers for their own graph, they are in the 2.0 business, whatever the deck says.
Dollar-survival figures are Truthset estimates, informed by CIMM identity benchmarking research (commissioned by Truthset).