A match rate tells you whether your data connected to an identity graph. It can't tell you whether you connected to the right people, how those connections were made, or how much of that audience you can reach.
Your identity provider ran your customer file and came back with a 90% match rate.
It's a good number, it's the number that gets screenshotted into the QBR deck, the number that helps justify the renewal, and the number nobody in the room questions because questioning it can feel like questioning the whole relationship.
Then the campaign runs, and the reach doesn't look like 90% of anything…
That gap between the identity match rate on the slide and the performance in the field is one of the most common frustrations in identity resolution. And it usually isn't because the number is wrong, it's because a single percentage is being asked to carry far more meaning than it was ever built to hold.
A match rate can tell you whether your data is connected to an identity graph. What it can't tell you on its own is whether you connected to the right people, how those connections were made, or how much of that matched audience you can actually reach.
Those distinctions matter.
What a match rate actually measures
A match rate tells you what proportion of the records in your file were connected to something in a provider's identity graph. That's it. That's the whole claim!
It doesn't tell you what they were connected to, it doesn't tell you how confident the connection is. It doesn't tell you whether the resulting audience can be activated anywhere useful. A 90% match rate and a 60% match rate can produce identical campaign outcomes, and in some cases the lower number performs better, because a smaller set of high-confidence person-level matches will outreach a larger set of loose household or device associations every time.
Two providers can also run the same file and report wildly different numbers, not because one has better data, but because they defined “match” differently. There's no industry standard forcing them to agree.
So the percentage isn't wrong, it's just incomplete in ways that only become visible after the money is spent.
The four questions a match rate doesn't answer
If you're evaluating a match report, these are the gaps worth closing before you act on the number:
- Was the match to a person or a household?This is the single largest source of inflation in reported match rates.
- Which identifiers carried the match?A match on a verified name and postal address is a different asset than a match on a shared device or an IP address that fifty people passed through.
- How much of the matched audience is actually addressable?Matched and addressable are not the same population, and the distance between them is where most of the disappointment lives.
- What happens between the report and activation?Every hop costs you something, and the number you were shown describes a moment that happened before any of them.
1. Was the match to a person or a household?
This is the single largest source of inflation in reported match rates.
Household-level matching connects your record to an address or a family unit. That's genuinely useful for some channels, direct mail, addressable TV, anything where the household is the buying unit. It's much less useful when you're trying to reach one specific decision-maker, and it becomes actively misleading when person-level and household-level matches are pooled into one blended percentage.
Ask for the split: A provider that resolves deterministically to individuals should be able to give it to you without a discovery call.
2. Which identifiers carried the match?
A match made on a verified name and postal address is a different asset than a match made on a shared device or an IP address that fifty people passed through. Both count as a match, only one of them survives contact with a measurement team.
If your report doesn't break down which identifier types contributed, you can't distinguish the durable portion of your audience from the portion that will decay before the campaign ends.
3. How much of the matched audience is actually addressable?
Matched and addressable are not the same population, and the distance between them is where most of the disappointment lives.
A record can be matched in the graph and still be unreachable in the channel you care about, because the identifier that carried the match doesn't exist in that destination. Your 90% match becomes a much smaller activatable audience the moment it hits a platform, and unless the report separates the two figures, nobody finds out until delivery reports come back.
4. What happens between the report and activation?
Every hop costs you something: If the ID you receive is encoded to work only inside one partner or environment, moving that audience elsewhere requires a translation. That translation can introduce match loss, latency, another vendor dependency and, depending on the provider, another line item, none of which shows up in the original percentage.
The number you were shown describes a moment that happened before any of this.
Why this became normal
None of this is a conspiracy, it's the predictable result of an industry that grew by acquisition.
Portfolios expanded, platforms multiplied, and new technologies had to coexist with legacy systems that measured things differently. Reporting layers got stacked on top of reporting layers. The percentage survived because it was the one output everyone could agree to show a customer, even as the machinery beneath it became something no single person at the vendor could fully explain.
Complexity accumulated, the number stayed simple and the customer absorbed the difference.
How to audit your match report
You need to ask harder questions of the one you have.
- What percentage of matches is resolved to an individual, and what percentage to a household?
- Which identifier types contributed to the match, and in what proportion?
- What is the addressable audience size for each destination we activate into, as distinct from the matched total?
- How is confidence scored, and can we see matches broken out by confidence tier?
- How current is the underlying data, and how often is it refreshed?
- If we move this audience to another partner, what does that require, what does it cost, and what match loss should we expect?
The answers matter, so does the response!
A provider that can answer these quickly and in writing is one whose numbers you can build a plan on. A provider that treats the questions as an unusual request has told you something useful about how much of your identity operation you actually have visibility into.
What transparency should look like
Identity resolution could be complex, and it's going to stay that way, nobody reasonably expects a vendor to publish its matching logic.
But complexity shouldn't require blind trust. There's a meaningful difference between proprietary technology and an unexplained result, and customers should be able to see what went in, what came out, what each metric represents, and what they're paying for at each step. That's not a demand, it's the minimum required to make an informed decision about your own data.
At Deep Sync, we use deterministic identity to connect data back to real people, households, and audiences, and we report those connections separately rather than blending them into one number. We don't encode IDs by default, so audiences stay usable across the ecosystem you choose. When a privacy or business requirement calls for encoding, we support it.
Sophistication should reduce the complexity you're managing, not relocate it onto your team.
Run the test yourself
The fastest way to find out what a match report should look like is to compare one.
Send us a file. We'll run a free match test and return a report showing what matched, what it matched to, which identifiers carried it, and how much of that audience is addressable in the destinations you actually use. No encoding requirement, no obligation, and the report is yours either way.
Then put the two reports side by side and see which one answers more of the six questions above.
Send us a file. We'll show you what a match report should look like.
We run a free match test and return a report showing what matched, what it matched to, which identifiers carried it, and how much of that audience is addressable in the destinations you actually use. No encoding requirement, no obligation, and the report is yours either way.
Julia is Director of Product Marketing at Deep Sync, specializing in identity resolution, customer data, audience activation, and go-to-market strategy. She helps brands turn complex data into actionable insights that power better marketing, personalization, and customer engagement.
Let's connect →
