Identity Resolution API: How It Works for Marketers
Send fragmentary customer data in, get one verified person back. What the call actually does, and five questions worth asking before you pick a vendor.
An identity resolution API is an endpoint you send fragmentary customer data to (an email, a hashed email, a phone number, a device ID, a name and postal address) and it returns a single persistent identifier for one real person, plus whatever attributes that person's record holds. It exists so your systems stop counting five partial records as five different customers. You can call it live from a form, a checkout, or an onboarding step, or run it in batch against a list you already own. That's the whole mechanism. The rest of this is how to tell a good one from an expensive one.
The part most teams haven't priced in yet
The matching most growth teams still run on: probabilistic stitching across device graphs, third-party cookies, and vendor scoring nobody on your team can audit, is on a timer. It works well enough today that rebuilding around it never makes the quarterly roadmap, and that's the trap. By the time it degrades past usefulness, the teams that already moved to deterministic, consent-backed resolution will have years of clean identity history behind them. Everyone else starts over in a hurry, in a quarter when performance is already slipping.
You don't have to buy that timeline. You just need a resolution layer you can explain to a privacy reviewer without flinching: which is the same layer that survives whatever replaces the current one.
New to the category? Read what is identity resolution first. This post assumes you're already choosing between providers.
What actually happens on the call
The shape is simple enough that vendor marketing tends to overdress it.
You send identifiers: some strong (email, phone, full postal address), some weak on their own (a first name, an IP, a truncated device ID). The API normalizes them: lowercasing, stripping plus-addressing, standardizing street suffixes, hashing to whatever format it expects. Then it looks for the person they point at.
You get back a confident match with a persistent ID, a partial match carrying a confidence signal, or nothing. That third outcome matters more than people expect. A provider that always returns something isn't more capable than one that sometimes returns null: it's less honest about uncertainty, and you pay for that in wasted spend.
Exact Match's own resolution tools (entity_resolve and entity_enrich) accept name, email, phone, and address as identity anchors and match deterministically against them rather than scoring a probability. More anchors supplied means more to verify against, not more noise to average over. Full endpoint list in the data enrichment API breakdown.
Deterministic and probabilistic, without the hand-waving
Deterministic matching links two records because they share something that can only belong to one person: a verified email, a phone number, a hashed identifier the person handed over themselves. Probabilistic matching links them because they behave alike, same household IP, similar browsing pattern, same device fingerprint.
Probabilistic isn't worthless. It's just not evidence. And it degrades in exactly the conditions the industry keeps moving toward: more consent gates, more identifier restrictions, fewer signals shared across sites. Deterministic data gets harder to collect and more valuable per record at the same time. Building a five-year asset? Build it from the material that doesn't rot.
Five questions to ask before you sign anything
1. What does your match rate actually count? This is the one that gets people. A 92% match rate can mean 92% of submitted records got any response, or 92% got a person-level ID, or 92% of a hand-picked demo sample matched. Three different products, same number on the slide. Ask what the denominator is, then ask them to run your file.
2. Where does the data come from, and what consent came with it? You're inheriting their sourcing decisions. Get provenance in writing before legal asks.
3. How fresh is it, and how often does it refresh? A record that was right eighteen months ago is a wrong answer delivered with confidence.
4. What happens on a no-match? Null, or a best guess dressed up as a match? Ask to see both response bodies.
5. How am I billed: per call, per match, or per record? Per-call pricing charges you for the misses; per-match aligns incentives better, though check whether a partial match bills as a full one.
Exact Match's own answers: no published match rate for resolution specifically (see the FAQ below for why); sourcing and opt-out handling documented in the customer data policy; the underlying graph refreshes daily; matching is deterministic, so a non-match returns nothing rather than a scored guess; and billing is none of the three options above: it's one flat Unlimited plan with unlimited credits, quoted on a consultation, so there's no per-call or per-match cost to optimize around at all.
Where this belongs in a marketing stack
Deduplicating what you already have. Your CRM holds the same person three times, under two emails and a typo. Resolution collapses them before you pay to email all three: see our deduplicate customer database workflow for the full process.
Cleaning lists before you spend. Suppressing dead records and merging duplicates is the cheapest performance improvement most teams have available, and nobody gets promoted for it, which is roughly why it never happens.
Enrichment. Once you know who someone is, you can attach attributes to them: a related but separate call. See data enrichment API for how the two fit together, and consumer data API for the attribute side.
Recognizing return traffic. Matching a known identifier back to someone you've seen before, so onboarding and suppression work off one record instead of a guess.
Our identity resolution solution page covers how these pieces connect end to end.
Run a sample before you run an integration
Don't start with engineering time. Take a set of records you already know the truth about (customers you've talked to, orders you've shipped) push them through, then read the misses by hand. That tells you more in an afternoon than a three-week pilot does, because you're measuring against ground truth instead of the vendor's scoring of itself.
Exact Match doesn't run a separate formal trial program, but the entry point is a demo rather than a sales cycle, and billing is monthly with no annual contract and 30 days' notice to cancel, so testing against your own file costs you a month or two, not a year, if it doesn't work out.
Frequently Asked Questions
What is an identity resolution API?
It's an endpoint that takes fragmentary identifiers (email, phone, device ID, name and address) and returns one persistent identifier for the real person behind them, plus any attributes on that record. Run it in real time from a form or checkout, or in batch across an existing list. The goal is to stop your systems treating one person's scattered records as several customers.
How is identity resolution different from data enrichment?
Resolution answers who this is, collapsing scattered records into one person. Enrichment answers what else you know about them, attaching demographic or behavioral attributes once identity is settled. Plenty of vendors sell both as one product, but they fail differently: bad resolution merges two people into one, while bad enrichment attaches a wrong fact to the right person.
Is using an identity resolution API legal under privacy regulations?
It can be, and the answer turns on sourcing and consent rather than the technique itself. What matters is where the underlying data came from, what people agreed to when it was collected, whether opt-outs propagate through the vendor's system, and whether you can produce that chain on request. Ask for that documentation in writing first, or read a vendor's own customer data policy before you sign anything. The same sourcing/consent question applies once that resolved identity feeds a downstream signal like intent, see buyer intent signals api. For how the same rules apply when matching without browser cookies, see cookieless identity resolution.
What match rate should I expect?
No honest vendor can tell you without seeing your file. Rates swing widely with data quality, record age, geography, and which identifiers you can supply. Treat any headline number as marketing until you've run your own sample and read the misses by hand, and confirm what the number counts before comparing vendors on it. Exact Match doesn't publish a match rate for resolution specifically; the 25-40% figure on our Site ID page measures verified human visitor identification, a different metric, not a substitute for this one.
Get Started: Unlimited
One plan, everything included: every product, every feature, and unlimited credits. Schedule a consultation and we will build pricing around your needs.