Subscribe to our newsletter
Get the latest insights on identity, data, and audience activation.
Identity resolution connects anonymous website activity to the people and households behind it, and every system that does it makes the same tradeoff. Match strictly, and the results are accurate but cover a small share of visitors. Match loosely, and coverage grows along with duplicate records and wrong matches.
It's natural to judge identity resolution by how many visitors it identifies, since a visitor you can't recognize is one you can't reach or measure. That number leaves out how many of those matches are right, and that's the half of the tradeoff that decides whether the data is usable.
What Limits Identity Resolution Accuracy and Reach?
Identity resolution uses two matching methods, and which one applies depends on what a visitor has shared. Each is strong where the other is weak.
Deterministic matching uses identifiers a visitor provides directly: a login, an email address, a phone number, or a completed form. They're high-confidence matches because the visitor has identified themselves. A click through from your email is the clearest example.
Probabilistic matching handles the visits that arrive without a direct identifier, like a return trip to your site days after that email click. It combines weaker signals such as device type, IP address, browsing patterns, location, and cookies to associate activity with a likely match. The more of those signals that line up, the more confident the match gets.
Why Deterministic or Probabilistic Matching Alone Limits Identity Resolution
A deterministic-only approach keeps the data accurate, since every match traces back to something the visitor confirmed. The cost is coverage. Many visitors browse without logging in or submitting anything. Deterministic matching can't recognize them, so the pool of people you can reach stays small.
Leaning on probabilistic matching expands that pool. It can also introduce duplicate records, false positives, and incorrect merges, where separate people get combined into a single profile. Probabilistic methods may improve match rates, but they often reduce overall data accuracy when nobody's measuring it.
Getting both accuracy and reach depends on how the two methods are combined.
How Your Use Case Sets the Right Identity Resolution Balance
Some providers rely on one matching method, as if every business has the same needs. Effective identity resolution blends the two: deterministic signals anchor resolution where a visitor has identified themselves, and probabilistic methods extend it to activity that would otherwise go unrecognized.
Where to set that balance depends on your business model and your audience's behavior, but most of all on what the data is for. Attribution that gets reconciled against sales records can tolerate some false positives in exchange for scale, because every match is checked against a real outcome later. Retargeting, email, and SMS are different, because a wrong match sends your message to the wrong person. Those channels need accuracy to clear a minimum bar first, and only then is more reach worth having.
The same business can need both settings at once. In automotive retail, for example, a dealer group measuring which channels brought in buyers can work with looser matching than the same group sending service offers to the households it has matched.
How to Evaluate an Identity Resolution Match Rate
Many vendors use one-size-fits-all settings and offer limited visibility into match rates, confidence levels, or overall precision. That makes the match rate the number most likely to be quoted and the one that needs the most scrutiny.
A match rate is the share of visitors a system resolved. Accuracy is the share of those matches that are correct. A system can push its match rate up by accepting weaker signals, and accuracy often drops when it does. Evaluating a match rate means measuring both.
Accuracy only means something when it's checked against identities that are already confirmed. So when a vendor quotes a match rate, the useful follow-up is what it was measured against, and whether you can see how accuracy moves as the settings change.
Our white paper walks through how we measure accuracy and use it to tune resolution settings.
How FocalGraph Balances Accuracy and Reach
FocalGraph adds probabilistic identity resolution to the deterministic identity your CDP already holds, resolving activity at the household level. It gives you control over the key resolution parameters, backed by measurable performance data, so with every adjustment you can see how the change affects match rates, attribution, and data quality.
The goal is matches you can trust, at the scale your use case needs.
