Every identity platform pitch eventually arrives at the match rate slide. Eighty-three percent match. Ninety-one percent match. The number is large, it sounds like precision, and it travels well in decks. The problem is that a high match rate is a measure of database overlap, not audience quality. Treating it as a performance signal is one of the more costly habits in programmatic media buying.
What Match Rate Actually Measures
When you upload a CRM file to an identity resolution platform or a DSP's onboarding tool, the match rate reflects the percentage of your records that the platform can link to a known identifier in its graph—a cookie, a device ID, a hashed email, or a resolved household. A 90% match rate means nine out of ten of your uploaded records found a corresponding node in that graph.
That is genuinely useful information about graph coverage. It is not information about whether those matched individuals are reachable through paid media, whether they are still in-market, whether the identity link is accurate, or whether the resulting audience will produce any measurable business outcome. A matched record could belong to a person who has opted out of targeted advertising, a device that hasn't been active in eight months, or an identity cluster that collapsed three different people into one profile.
The Gap Between Matched and Addressable
Addressability is the layer that matters for campaign delivery, and it sits downstream of matching. An identity may be matched but not addressable if the associated device is outside a walled garden's reach, if consent signals have excluded it from targeting in regulated environments, or if the publisher inventory where that ID is active doesn't intersect with your media plan.
In practice, the delta between matched records and truly addressable, deliverable impressions can be substantial. Buyers who don't interrogate this gap often discover it only at the end of a flight, when reach was far lower than the match rate implied. By then, frequency has been compressed against a smaller pool, and cost-per-reachable-person is meaningfully higher than modeled.
A clean diagnostic habit: after any audience activation, pull the delivered reach number and divide it by the uploaded file size. That ratio—call it your effective addressability rate—is more operationally honest than the match rate you received at onboarding.
Identity Accuracy Is a Separate Problem
Even when a record is matched and addressable, identity accuracy determines whether you're reaching the right person. Identity graphs are probabilistic by design. They make inferences from behavioral signals, device co-location, login data, and purchase patterns. Those inferences degrade over time as people change devices, move, or share hardware.
A match rate of 90% with a graph accuracy rate of 70% means roughly 63% of your audience is both matched and correctly identified. That's a meaningfully different planning input than 90%. And graph accuracy is rarely disclosed in a standardized way by platform vendors, which means buyers have to proxy it through outcome data.
One practical proxy: run a small holdout test against a known customer segment where you have ground truth—recent purchasers, for example—and measure onsite conversion or CRM match-back against delivered impressions. If the audience was accurately resolved, conversion signals should align with expectation. If they don't, identity accuracy is a likely culprit.
Better Metrics to Build Into Your Evaluation Framework
Replacing match rate with a richer set of signals doesn't require a new vendor or a complex integration. Most of the relevant data is already available in your DSP reporting, your measurement platform, and your CRM.
Effective reach rate. Delivered unique reach divided by uploaded audience size. Tracks how much of your intended audience actually received an impression.
Frequency distribution by identity tier. If your DSP supports it, segment delivery by identity confidence tier. Heavy delivery against low-confidence IDs is a signal that your addressable pool was thinner than expected and the algorithm back-filled.
Audience overlap across platforms. If you're activating the same seed audience across two or more DSPs and match-back shows minimal overlap in the people actually reached, your identity graphs are diverging significantly. That's a data quality signal, not just a reach story.
Suppression effectiveness. If you're uploading a suppression list—recent converters, existing subscribers—measure how often those individuals still appear in delivery logs. Suppression leakage is a direct test of identity resolution accuracy and often reveals graph fragmentation you wouldn't otherwise see.
CRM match-back rate post-campaign. What percentage of reached devices ultimately match back to known customer records in your CRM? Low match-back rates suggest the audience you reached and the audience you intended to reach drifted apart during activation.
A Word on Graph Selection
None of this is an argument against using identity resolution platforms. The ability to activate first-party data against scaled paid media inventory is genuinely valuable, and the infrastructure that enables it is sophisticated. The argument is for evaluating that infrastructure with the same rigor you'd apply to any other media investment.
When comparing graph providers or activation partners, ask for transparency into graph construction methodology, recency of the underlying signals, and how the platform handles identity in cookieless environments or privacy-regulated contexts. Vendors who answer those questions specifically are worth more scrutiny than vendors who lead with match rate percentages.
Also consider what happens to your data in transit. Identity resolution involves hashing, matching, and in some architectures, enrichment against third-party data. Understanding data handling and contractual protections around your first-party file is a procurement discipline, not a legal afterthought.
The Operational Shift
The practical change here is small but consequential: move match rate from the evaluation stage to the baseline stage. Confirm that graph coverage meets a minimum threshold for your use case, then shift attention to the metrics that predict campaign performance—addressability, accuracy proxies, and outcome alignment.
Media buyers who make this shift stop optimizing for impressions against matched records and start optimizing for impressions against reachable, correctly identified people. That distinction compounds over time into meaningfully better ROAS, lower wasted spend, and audience strategies that hold up when you actually look at the data.