Every marketer I talk to is running some version of the same experiment. They’ve spent three or four years building a first-party data asset — a clean CRM, a consented email file, a loyalty program, purchase history tied to real customers. It’s the most valuable thing the marketing organization owns. And then they go to spend media against it, and somewhere between the audience build and the attribution report, the thread goes slack.
The audience matches at a rate someone else calculated. The impressions deliver into environments where the identifier they built their business on doesn’t exist. The conversions come back as a number in a dashboard with no way to interrogate how that number was produced. The asset is real. The connection to outcomes is a leap of faith.
That gap is usually described as a measurement problem. It isn’t. It’s an identity problem wearing a measurement costume.
The signal doesn’t disappear. It changes shape.
Here’s the mechanical version of what’s happening.
A brand’s first-party data is organized around a person: an email address, hashed for safety, sitting in a CRM. That’s the spine of everything downstream — the audience, the suppression list, the LTV model, the conversion event.
But a growing share of media doesn’t deliver against a person. It delivers against a device, a household, a connected TV, a session. Connected TV is the clearest case: no cookie, no reliable app-level identifier in many environments, and an audience sitting on a couch rather than logged into anything. The signal that’s actually present in that impression is an IP address. Same for a meaningful share of programmatic display, and same for any environment where the browser has quietly stopped carrying the identifier the industry spent twenty years building around.
So the brand has a hashed email. The media has an IP. Those two things describe the same customer, and nothing in the standard stack connects them in a way the brand can see, test, or explain.
What fills that gap today is usually somebody else’s graph. You hand over your audience, it comes back matched, and you’re told what the match rate was. You can’t see the logic. You can’t tell whether a match is two weeks old or two years old, whether it’s built on a stable residential connection or a mobile carrier IP that rotated three times this morning. You can’t tell whether the “match” is one strong relationship or four weak ones averaged into something that looks confident.
That’s the black box. And it isn’t just an intellectual irritation — it has real downstream consequences. When you can’t see the quality of an identity link, you can’t tell the difference between a campaign that underperformed and a campaign that was measured against bad connective tissue. You optimize against noise. You defend budget with a number you didn’t produce and can’t reproduce.
What we built, and why this pathway specifically
Matchbook is an identity augmentation platform. Our job is not to be your identity graph — it’s to make yours work harder by connecting the identifiers you already hold to the signals actually present in the environments where you buy media.
We’ve just launched a new data matching pathway connecting IP addresses to hashed emails (HEM), now available in the US market.
Practically, that means a brand, agency, or ad tech platform can take the hashed email file they already own and connect it to IP-delivered inventory — for reach, for retargeting precision, and, most importantly, for closing the loop between an impression delivered against an IP and a purchase that happened under an email address.
This sits alongside the identifiers Matchbook already supports — lat/long coordinates, MAIDs — rather than replacing them. That combination matters more than any single pathway. One of the quieter costs in this industry is fragmentation: identity resolution stitched together across three vendors, each with its own delivery format, its own refresh cadence, its own scoring approach, and no common frame for comparing any of it. Consolidating core identifiers in one place isn’t a convenience feature. It’s what makes cross-channel measurement coherent instead of a reconciliation exercise.
For CTV in particular, this is the difference between “we reached a household” and “we reached a household that subsequently converted online, and here is the identity link that supports that claim.”
Confidence scores as a transparency instrument
The part of this launch I’d argue most about, if anyone wanted to, is the Confidence Score.
Every IP-to-HEM relationship carries one. It’s dynamic — it continuously evaluates and measures the strength of that relationship rather than freezing a judgment made at the moment of ingestion. IP-to-person relationships are not static. Households move. Connections change. A link that was strong in March may be weak by September, and a system that doesn’t account for that is quietly selling decay as certainty.
The reason I want to frame this as a transparency feature rather than an accuracy feature is that accuracy claims in identity are close to unfalsifiable. There’s no universal truth set. A single accuracy percentage for an identity graph is a number produced by a methodology you can’t inspect, against a benchmark you didn’t choose. The more useful version of the conversation isn’t “trust our number.” It’s “here is the strength of each link, here is how we assess it, here’s how it moves over time — set your own threshold.”
That’s a different posture toward the buyer. It puts the tuning dial in your hands. Want maximum precision for a high-value retargeting segment? Set a high threshold and accept smaller scale. Building a reach campaign where the cost of a loose match is low? Open it up. Those are business decisions, and they should be made by the person accountable for the outcome, not embedded in a default setting and never surfaced.
It also changes what an attribution report means. A conversion tied to a high-confidence identity link and a conversion tied to a marginal one are not the same evidence, and collapsing them into one number is how measurement loses credibility with finance.
Privacy-forward isn’t a caveat paragraph
One more thing, because it usually gets relegated to the last line of a press release.
This pathway is built on hashed emails. The design intent is to bridge signal gaps without moving raw personal information around, and without asking brands to trade consumer trust for scale. As Vinod Kashyap, our Chief Product Officer, put it: the goal is a privacy-forward foundation for first-party data owners that bridges signal gaps without sacrificing accuracy and scale.
What I’d add is that privacy and transparency are the same discipline pointed in two directions. A system that can explain how it resolved an identity is easier to trust, easier to document, and easier to bring to a partner who has their own standards to meet. Confidence-scored, inspectable identity linkage is better practice on both counts — not as a defensive measure, but because it’s how you build infrastructure that holds up as the ecosystem keeps changing.
The ask
The industry spent the last several years telling marketers to build first-party data. That was right. The unfinished half of the instruction is that first-party data only creates value at the point where it connects to media and to outcomes — and that connection is currently the least transparent link in the entire chain.
So the question worth putting to every identity partner, including us: show me how you decided these two things are the same person, and show me how confident you are today.
If the answer is a match rate and a shrug, that’s your black box.
Matchbook IP to HEM is available now in the US. More at matchbookdata.com.
