Forty cents of every dollar spent in the open CTV programmatic market this year gets wasted. Not on fraud. Not on wasted creative. On bad identity data. That is the headline number out of Truthset’s 2026 State of Data Accuracy report: roughly $7.4 billion in lost CTV value, in a channel eMarketer expects to hit $38 billion in U.S. spend this year, with 88% of that already moving programmatically.
I have spent fifteen years building identity infrastructure for this industry, and this number does not surprise me. It confirms what anyone running audience matching at scale already suspects. CTV’s real bottleneck was never inventory fragmentation. It is identity.
The Inventory Story Was Always Incomplete
The standard complaint about CTV is fragmentation, too many apps, too many walled gardens, a single household showing up as ten different “unique” viewers across ten different streaming services. FreeWheel’s Alex Ibarguen said it plainly earlier this year: CTV fragmentation is very much an identity problem. He is right, and the inventory framing has let the industry avoid the harder conversation for too long.
CTV has no cookie to lean on, imperfect as that crutch always was. So buyers piece together IP to household mapping, probabilistic device graphs, and inference to figure out who they are actually reaching. Every one of those links introduces error, and the errors compound. Truthset traces it start to finish: bad IP to household connections lead to misclassified demographics, which break attribution, which sends media to the wrong audience entirely.
Where the Accuracy Tax Actually Lands
Advertisers pay it when their targeting misses the intended household. Measurement platforms pay it when they log outcomes that never happened. Media teams pay it when they have to explain to leadership why performance does not match the reported reach.
The Coalition for Innovative Media Measurement put numbers behind this in its Identity Infrastructure 2.0 research. Linkage accuracy on hashed email to postal matches averages around 51%. IP to household inference, the backbone of most CTV targeting, does worse. CIMM’s read is that deterministic identifiers, the ones anchored to verified real world data instead of statistical guesswork, are coming back as the stabilizing layer this ecosystem needs. Interoperability across identity providers is no longer optional. It is becoming the price of entry for anyone who wants CTV dollars to work as hard as linear once did.
Reach Was the Old Pitch. Relevance Is the New One.
Lotame’s Chris Hogg framed the shift well: in 2026, CTV moves from reach to relevance, and identity resolution plus interoperability, not raw impression counts, is what actually drives audience connections that move a business outcome. For most of CTV’s growth run, the pitch was scale, more inventory, more screens, more reach than linear could ever deliver. That pitch now needs proof the audience on the other end is real, correctly classified, and reachable again for frequency management and measurement.
That is why data quality and identity vendors are moving closer together instead of staying in separate lanes. Truthset has spent the past several months wiring its validation layer directly into Databricks, Magnite, and Unified ID 2.0. Experian rolled out a data marketplace meant to unify identity, interoperability, and addressability across CTV, display, and mobile in one activation layer. None of this is coincidence. Identity infrastructure and data accuracy are becoming the same problem, because you cannot fix one without the other.
What This Means If You Build or Buy Identity
A few things follow from this, and they apply whether you are building the graph or buying against it.
Validated data matters more, not less, in a channel this heavy on probabilistic modeling. Every MAID, hashed email, or household IP anchored to verified, first party sourced identity meaningfully improves match quality downstream, even if it only covers a portion of the graph.
Interoperability is turning into a purchasing criterion on its own. Buyers increasingly want to know whether an identity partner plugs into the clean room, measurement, and activation stack they already run, not whether they can be talked into rebuilding around one proprietary graph.
And validation stops being optional. The 40% waste number exists because the industry spent years activating audience segments without independently checking whether they were accurate. Expect more RFPs to ask vendors directly how their match rates and demographic classifications have actually been benchmarked, and against what.
The Bottom Line
CTV is not shrinking. The eMarketer numbers make that clear, and the format still delivers something linear never could, big screen storytelling with digital level measurement attached. But that promise only holds if the identity underneath it can be trusted. We built BDEX’s identity graph around this exact problem, tying MAIDs, device IDs, and household level data back to verified sources instead of pure inference. Whether you work with us or not, the lesson from this report is the same for everyone in this industry. Identity is not a line item under measurement anymore. It is the infrastructure everything else in CTV sits on top of.
