Match rate is the percentage of the records you send to a data provider or an ad platform that it can tie to a record of its own and return something for. Send 10,000 valid contacts to an enrichment tool, get data back on 7,000 of them, and your match rate is 70%.

Two groups use the word, and they mean different things by it. People running data enrichment use it for how much of a list a provider recognized. Paid media teams use it for the share of an uploaded customer list that LinkedIn, Meta or Google could link to their own users. The arithmetic is identical. What it tells you, and what you can do about a low one, is not.

Either way the number gets more trust than it has earned. A match only says the provider found a record it believes is yours. It doesn't say the record is current or that it's even the right person, and the match rate on a vendor's sales deck was measured on a list that wasn't yours.

TL;DR

• Match rate is matched records divided by valid records sent. Decide what counts as valid before the upload, because duplicates and rows with no usable identifier pull the number down without telling you anything about the provider.

• Fill rate is a separate number. A matched contact can still come back without the phone number or title you needed, so track the rate for the field your play depends on, next to the match rate.

• Ad platforms play by their own rules. LinkedIn only matches members with a verified email and shows "under 5%" until 300 members match, Google shows a rate only once at least 100 rows match, and Meta and Google both compare SHA-256 hashed identifiers rather than raw ones.

• Your input decides most of the result. A company domain or a work email matches far more reliably than a name, which is why HubSpot's own enrichment works only from first name, last name, work email and company domain.

• Matched isn't the same as right. A provider matching 90% of a list at 60% accuracy leaves you 54% usable records, while one matching 70% at 92% accuracy leaves about 64%.

• Treat any vendor's match rate as a claim made on someone else's list. The figure worth planning around comes from a test on records you already know, split by segment.

• A waterfall of providers lifts the combined rate because each source plugs part of the hole the previous one left, but every extra source adds less than the one before it.

How to calculate match rate without flattering or punishing the provider

The formula is matched records divided by valid records submitted, times 100. Most arguments about a match rate are really about the bottom half of that fraction.

A Rotterdam freight software company exports 10,600 contacts from its CRM to test a new provider. Four hundred rows are duplicates, and another 200 have neither an email address nor a company domain, so no provider could ever find them. That leaves 10,000 valid records. The provider returns data on 7,000, which is a 70% match rate. Divide by the raw export instead and you get 66%, a lower number that only measures how messy the export was.

The opposite mistake flatters the provider. If you drop every record without a work email before the test, the hard cases never reach the denominator and the rate looks better than it will on a real list. So write the rule for "valid" down before the first upload, apply it the same way to every provider you compare, and do the obvious data hygiene (duplicates, empty rows, test entries) first.

Then agree on what counts as a match. Some tools report a record as matched when they found the company but not the person, or when they returned a single field you didn't ask for. For a contact list, we'd only count a record as matched when the provider returned the person, not just their employer.

Match rate, fill rate and the field you actually needed

A matched record isn't necessarily a complete one. The provider can recognize Anna at Nordwind and still have no phone number for her, which is where fill rate comes in. Match rate counts records that came back with something. Fill rate counts, field by field, how often the thing you wanted was in there.

Back in Rotterdam, suppose 2,900 of the 7,000 matched contacts came back with a direct dial. The match rate is still 70%, but the phone fill rate on matched records is about 41%, and on the full valid list it's 29%. If the team behind that test runs a calling motion, 29% is the number that decides whether the provider is worth paying for.

You'll also see coverage rate used for much the same idea as match rate. Vendors swap the two freely, so ask which one a figure refers to and how it was counted, rather than assuming.

Ad-platform match rates follow different rules

When marketing uploads a customer or target-account list to an ad platform, the platform is matching your rows against its own users, not against a B2B database. Each platform does it differently, and the reporting catches out people used to enrichment tools.

The details come from the platforms' own documentation. LinkedIn's help page on match rates defines the rate as the share of entries that matched member accounts within the past seven days and warns that contact targeting only matches members with verified email addresses. Meta's Custom Audiences documentation sets out the hashing and normalization rules and says that sending several keys per record increases the match rate. Google's page on Customer Match match rates adds a line worth remembering: the rate isn't an indicator of how the list will perform.

B2B lists tend to struggle on the consumer platforms. A file of work addresses is matched against accounts that many people opened with a personal email or a phone number, so a decent CRM export can come back with a low rate on Meta or Google and a much better one on LinkedIn. Adding phone numbers and names you already hold gives the platforms more to work with. Buying personal emails just to lift an ad match rate is a different decision with legal weight in Europe, and it's rarely worth it.

The 300 figure turns up again when you build on a list. LinkedIn's predictive audiences need at least 300 matched member accounts behind the seed, so a small, tightly defined account list can be too thin to build from even when every person on it is exactly right. What LinkedIn does with a seed that clears the bar is covered under B2B lookalike audiences.

What you send decides most of your match rate

Every lookup starts from the identifiers you supply, and some are far easier to find than others. A company domain points to exactly one business. A work email usually points to exactly one person. A first name and a company name point to a shortlist, and a name alone points to almost anyone. How confident each kind of key can make you is covered in more depth under identity resolution. For match rate, the practical upshot is that the same provider will post very different numbers on two lists depending on which keys they carry.

Typos do more damage than people expect. A misspelled email, a company name with a stray legal suffix ("Nordwind GmbH" in one system, "Nordwind" in another), or a freemail address standing in for a work one all come back empty, and none of that is the provider's fault. Standardizing the fields through data normalization and running addresses through email validation before a test removes most of it. Filling in a domain from your firmographic data where you only have a company name often helps more than switching providers.

Companies and contacts also behave differently, so report them separately. Company matching keyed on a domain tends to run well ahead of contact matching, because a business keeps its domain for years while people change jobs, titles and addresses. A provider can be excellent at firmographics and technographic data and mediocre at finding the right person inside each account, and a blended figure hides which is which.

A matched record can still be the wrong record

A match is the provider's belief that its record and yours describe the same entity. Sometimes that belief is wrong. Two people with the same name work at the same company. The contact left two years ago and the provider never noticed. A subsidiary's employees get matched to the parent company's domain. Each of those counts toward the match rate, and each one sends a rep to the wrong person.

The fix is to read match rate next to accuracy. On 10,000 valid records, a provider that matches 90% with 60% of those matches correct leaves you 5,400 usable records. One that matches 70% at 92% accuracy leaves 6,440, about 64% of the list. If both bill €0.20 per lookup, the €2,000 test costs about €0.37 per usable record with the first provider and €0.31 with the second. The provider with the smaller headline number is the better buy, and the same comparison sits at the heart of any trial when you're choosing a data provider.

Accuracy needs its own checks, separate from matching. Run returned emails through contact verification, check a handful of returned job titles by hand, and remember that even a correct match starts aging the day it lands because of data decay. A match rate measured last year says little about the records you're emailing this month.

Reading a vendor's match rate claim

Most published match rates come from vendors, and they're measured the way the vendor chose: on a list the vendor picked, with keys the vendor picked, under a definition of "match" the vendor picked. That doesn't make them false, but it does make them marketing. Before one goes into a business case, ask a few questions.

• Whose list was it measured on, and does it look like yours in country, company size and seniority?

• Which keys went in? A test run on clean work emails will beat one run on names and company names every time.

• What was the denominator: everything uploaded, or only the records the vendor considered valid?

• Does a company-only result count as a match on a contact list?

• Is it the rate for any field returned, or for the field you're buying (a verified email, a mobile number)?

• Were the matches checked for accuracy, and when was the test run?

Then measure it yourself. Pull a few hundred records you know are correct, spread across the segments you sell to, cut them down to the identifiers you'd normally have, and run that one file through every candidate. Read the results by segment, since a 50% match rate on a hard segment can be a stronger result than 80% on an easy one. We don't publish a "good" match rate for B2B, because no figure we could find was measured on lists comparable to anyone else's.

Pushing match rate up with a second and third source

No single database covers a whole market, and the gaps differ from provider to provider. That's the reason to chain them. In waterfall enrichment, every record goes to the first provider, and only the misses move on to the next source down.

Picture a list where the first provider matches 60%. The second picks up another 15 points from the leftovers, taking the combined match rate to 75%, and the third adds 8 more for 83%. The pattern in that illustration is the one to expect: each source adds less than the one before, while it still costs money to ask. After a handful of sources, the extra records stop paying for the lookups.

Billing matters here. Some providers charge for every lookup, found or not, while others only charge when they return a result. In an email waterfall, Clay states in its waterfall lesson that it refunds credits when a provider doesn't return an email and stops once a valid one is found. Lead with the cheap sources that find a lot, and the pricey ones are only asked about the leftovers.

A higher combined rate is also what makes everything built on top of enrichment work at full strength, from list building to routing and scoring. And once you let prospecting run on a schedule, a few points of match rate add up to a lot more reachable contacts every month, with no extra work from the team.

Wrapping up

Match rate is a useful first question and a poor final answer. It tells you whether a provider or a platform can see the people you care about at all, and if it can't, nothing else about that source matters much.

The trouble starts when it gets read on its own. A rate depends on how you counted the denominator, which keys you sent, whether the matches were right and whose list it was measured on. Hold those steady and the number becomes comparable. Let them drift and you're comparing a vendor's best day with your own messy export.

The habits that fix most of it are unglamorous. Clean the list before you test, count companies and contacts apart, put accuracy and fill rate beside every match rate, and trust your own sample over anyone's slide. Those records then feed the scoring, routing and outreach that run on your data, so the care you take here shows up everywhere downstream.

Nebor runs match tests like this on records a client already trusts, then builds the chosen sequence of providers into the client's Clay workspace and CRM. The test file, the results and the reasoning stay with the client, ready to rerun when a provider changes its data or its prices.

Frequently asked questions