On 13 September 2026 we read 41 vendor pages that sell phone data to call centers. The ones that quote a match rate quote it as a single number, and not one of the 41 says what a match is. That is not because the vendors are hiding something — it is because the word has been doing four different jobs for so long that each vendor assumes its own meaning is the meaning, and a buyer who asks gets a slightly puzzled “the percentage of records we match.”
This page is the definition that the 41 pages leave out: the four things a match rate can count, what each one predicts about a dialer day, how to get any vendor to say which one they report, and how to read all four off your own dialer export so you never have to take the number on faith again.
The word does four jobs
A match rate is a fraction. The denominator is usually the records you sent — addresses, or addresses with names. The numerator is where the four meanings diverge, and the same pool of homes produces four very different figures depending on which is chosen. The numbers below are the ones on our benchmarks page, each with the sample it was counted on.
1. Coverage: a phone came back
Some number is now attached to the row. It may belong to the owner, a previous owner, a tenant, an adult child, or a landline that has not been answered since 2014. Coverage measures the breadth of the vendor’s identity graph and the looseness of its matching threshold, and nothing else. It is the easiest of the four to push above 90%, because returning something always counts, and it is what most published “95%+” figures are. Our skip tracing accuracy page calls it a measure of effort rather than usefulness, which is fair.
2. Owner match: the number belongs to the person on the deed
The county says who owns the house. The phone was resolved to that person, by name, not to the address. This is the first meaning that predicts anything about who answers, because a wrong-party connect is almost always an address-matched number reaching whoever the data last associated with the house. On an 18,309-parcel pool of single-family homes in suburban Dallas–Fort Worth, measured 18 August 2026, 79.7% of homes yielded a contact whose name matched the deed.
3. Active mobile: that owner has a live wireless line today
The owner is matched and the number is a mobile that shows carrier activity now, not a landline and not a line that was disconnected in the spring. On the same Texas pool, 54.6% of homes yielded that. The 25-point drop from the line above is the gap that a one-word “match rate” hides best, and it is the number a delivery file is actually built from: a row with no reachable phone is not a row.
4. Connect: somebody picked up
No vendor can put this on a rate card, because it happens on your dialer, on your hours, with your caller IDs. On a Bay Area solar floor’s predictive dialer, agent connects ran 5.02% per dial on the first pass over a 25,000-record file (24,879 dials, 1–2 September 2026); on a three-state solar floor, a live human answered 4.45% of 2,224,746 dials in the sixty days to 25 August 2026. The first three meanings are what a vendor did; this one is what the file did.
Line the four up on one pool and the shape is: a phone of some kind for most homes, the owner for about eight in ten, the owner on a live mobile for about half, and a pickup on roughly one dial in twenty. A vendor can pick any of the first three, print it, and be telling the truth. Which is why the question is never “what is your match rate” — it is “match of what, over what.”
Getting a vendor to say which one
Five questions, in order. Each has an answer that names the meaning without the vendor needing to know this vocabulary, and a good vendor answers all five in a minute. None of this assumes anyone is hiding anything; it assumes the word is ambiguous, which it is.
- “Match rate of what, per what?” If the answer is “records with a phone returned,” you have coverage. If it is “records where we tied a phone to the named owner,” you are at owner match. Anything vaguer than either is coverage.
- “Is the name on the phone record checked against the name on the deed?” A yes with a description of how — vesting names, trusts, initials — means the vendor has an owner-resolution step. A yes with no description usually means the address did the matching.
- “What share of the matched numbers are wireless, and what share showed carrier activity in the last twelve months?” Two numbers, or an admission that the vendor does not carry them. Either is useful. A file that cannot state its mobile share cannot be loaded into a dialer mode that depends on it.
- “When the owner has no reachable number, is the row dropped or filled?” Filled with a relative, a resident, or a landline is how coverage gets to 95%. Dropped is how a file gets to half. Both are legitimate products; only one of them is a dialing file.
- “Will you put that definition on the order?” A sentence — “match means a phone tied by name to the owner of record, wireless, active within twelve months” — is the difference between a claim you can test for a credit and a claim you cannot.
Reading all four off your dialer export
The one place every meaning becomes visible is the export your dialer produces after a shift. Vendors cannot see it and rarely ask for it, which is why it is the most under-used file in the building. The method below is the one we use to publish our own numbers; the full four-test audit built around it is how to audit a lead vendor.
- Export with the right columns. Dialed number, disposition, attempt number, timestamp, agent. Most platforms will add the list or campaign name; take it, because it is your source stamp.
- Join it back to the delivered file on phone number first, then on owner name. Every dial now sits next to the row that produced it. A batch that matches zero rows is a scoping mistake, not a finding.
- Code three outcomes, as three codes. Right party, wrong party, dead or non-residential. “Not interested” is a right-party outcome, not a fourth code; a voicemail is neither a connect nor a dead number.
- Compute the inverses. Dead as a share of dials is the inverse of active — how much of the file was never a live line. Wrong party as a share of connects is the inverse of owner match — how the file was matched. Connects as a share of dials is the fourth meaning itself.
- Compare at the same attempt depth. Connect falls with every pass over a file — 5.02%, 3.55% and 3.45% on the first, second and third dial of the same Bay Area file — while appointments did not (5, 3 and 6). A first-pass number against a third-pass number proves nothing about either vendor.
- Put your current file beside it. Same reps, same hours, same codes. A wrong-party share means nothing alone and everything against the list you are already paying for.
Two shifts of dials on a hundred rows will tell you which of the four meanings you bought. It will not tell you a vendor is wrong — nobody is wrong about a number they did not define — but it will tell you what the number was measuring, and whether to keep paying for it.
Which two we report, and why
Disclosure: Scout Data is our product, and this is the definition we sell to. A record, for us, is meanings two and three together — one home, the owner on title, a live mobile — which is why our yields read as roughly half of a pool rather than 95%: on the Texas pool above, the other 45.4% of homes were dropped, not filled with a relative or a landline. Connect, the fourth meaning, we publish only from customers’ dialer exports, with the sample and the month beside each figure. A coverage number is never presented as accuracy, because a floor cannot dial coverage.
That is a choice, not a law of nature, and vendors who tune for coverage are not wrong to; a mail house or a researcher wants the opposite trade. What a call center should refuse is the one-word version. Ask which of the four, get it in writing, and read the export. The pre-purchase check on a sample is walked through in homeowner phone lists for call centers, and the reason two floors quoting the same “contact rate” are usually dividing by different things is in what is a good contact rate.
Frequently asked questions
What is a good match rate in skip tracing?
It depends entirely on what is being counted. A rate above 90% is almost always coverage — a phone of some kind came back for the address. Measured as the owner on title with a live mobile, the same kind of pool lands much lower: on one 18,309-home Texas pool we measured in August 2026, 79.7% of homes yielded the owner and 54.6% yielded the owner with an active mobile. A vendor quoting the second or third figure is telling you more than one quoting the first.
Is match rate the same as hit rate?
Usually, and usually both mean coverage: the share of records for which the vendor returned something. Some vendors use “hit” for any return and reserve “match” for a return tied to the named person, but there is no convention, which is why the only safe move is to ask what the numerator is and get the answer on the order.
Why does a high match rate produce a low connect rate?
Because the two count different things at opposite ends of a funnel. A coverage match adds rows; it does not check that the number rings, that it is a mobile, or that the owner is on the other end. Each of those is a gate the row still has to pass, and a file tuned to maximise the first number tends to fail the later gates more often — which shows up on the floor as dead dials and “they moved.”
How many records do I need to check a match rate myself?
Fifty to a hundred rows you drew yourself will tell you which meaning you bought — twenty names against the county record and a shift of dials is enough to see whether the wrong-party share is a handful or a third. A number you would put in a contract wants a few hundred rows dialed on a normal shift, against your current file, at the same attempt depth.