What One Verification Attempt Established About 805 Commercial Phone Numbers

A single verification attempt (a call, together with supporting review where the call alone left identity open) against each of 805 phone numbers supplied with commercial collection placements, one number per business, testing what a first attempt establishes about a number's connection to the business named on the account.

The result

Important scope note. These results describe one placement population, measured on a single call attempt per number. They estimate neither the accuracy of commercial phone data generally nor the reason any individual number failed. No published dataset measuring verification rates on commercial contact information at placement was found, so the 37.4 percent figure has no external benchmark to sit against and should be read as a first measurement rather than a comparison.

JSD ran one verification attempt against each of 805 phone numbers supplied with commercial accounts at the point of third-party collection placement. An attempt consisted of a call, together with supporting review where the call alone left identity open. After that attempt, 301 numbers were verified as belonging to the business named on the account, 105 produced a definitive negative result, and 399 remained unresolved, nearly half the file finishing the attempt without establishing either a match or a mismatch.

The three-way split is the study. Of the 504 numbers that did not verify, 399, or 79.2 percent, remained unresolved rather than being definitively disproven. Unverified is an evidence state rather than a verdict on the data, and treating the 62.6 percent that failed to verify as bad numbers would overstate what a single attempt established.

How verification was defined

A number was classified as verified when the evidence from the call, together with supporting verification signals, was strong enough to tie it to the business expected on the account. Usually that came straight from the greeting, whether a receptionist, an auto-attendant, or a voicemail message naming the company. A smaller share needed corroborating information.

Because that judgment is part of the result, it is worth testing how far the finding moves under stricter definitions.

Of the 301 verified numbers, 255 were confirmed through a recording, meaning 227 where the recording named the business and 28 where an unnamed recording was paired with corroborating evidence. The remaining 46 were confirmed on a live answer.

An exact name match would have discarded half the good numbers

What separates the first and third rows of the table above is a coding rule, and the choice between those rules can determine whether a potentially usable contact record is recognized or rejected.

Among the 301 verified numbers, a name comparison matched 159 to the expected business exactly. The other 142, or 47.2 percent, answered under a name that did not match the account, and the reviewer was able to verify them anyway. A rule that accepted only exact name matches would have recognized 159 of the 301 numbers ultimately verified and thrown away the rest as unusable.

A greeting stating some other company name is weaker evidence than it first appears. In this type of review, a business may answer under a trade name, franchise entity, subsidiary, or operating company rather than the invoiced entity, and that entity may carry no obvious relationship to the name on the account. The pattern a reviewer worked from, using illustrative names:

The same rule guards the other direction. A similar or overlapping name was not treated as verified unless the business could be tied to the account through supporting evidence such as address, trade name, ownership, or operating-entity relationship.

Two different questions

Every call was coded on two dimensions, what physically answered and what the result established about identity. Those are separate variables, and they cross one another. This is a high-level roll-up; the full cross-tabulation appears in the methodology below.

No single endpoint settles identity. A voicemail greeting produced a verification 105 times and identified a different entity 52 times, while 112 voicemails settled nothing about identity at all.

What the call reached instead

Under this single-call protocol, 441 numbers, or 54.8 percent, neither verified the business nor put a person on the line.

Several of those look identical from a desk in accounts receivable, and none of them establishes that the business named on the account is avoiding contact.

Reaching a person

A live person answered 109 calls, or 13.5 percent. Of those, 46 were on numbers that verified as the expected business, 7 were tied to a different entity, and 56 stayed unresolved. Phone systems turned up far more often, with a tree or menu answering 231 calls, or 28.7 percent.

A live person answered on a number verified to the expected business in 46 cases, or 5.7 percent of the file. Phone trees carried 150 of the 301 verifications, against 46 that involved a person on the line. A line producing a recognizable business endpoint and a person being reached on the attempt were separate outcomes.

What this means for a creditor

An accounts-receivable team that has heard nothing back tends to read the silence as avoidance. The calls go out, nothing comes back, and the conclusion writes itself. This study tested one call per number and can say nothing about repeated attempts or about intent.

What it can say is that the same apparent silence corresponds to materially different conditions. A number may simply have rung without answer, reached a generic system whose owner could not be established, reached a different company, failed to route, or hit screening that never passed the call along.

Two of these findings have a practical edge. The first is the name comparison. Any process that decides whether a contact record is still good by matching the answering name against the account name, whether that process is a software rule or a collector working a list, would have discarded 142 numbers that turned out to be the right business. Establishing identity took a person reading a website and comparing an address, a step a name-matching rule alone does not perform. The second is the 5.7 percent figure. A verified line and a reachable person are different outcomes, so confirming that a number is verified confirms only that the line exists, not that a person can be reached through it.

Where the underlying uncertainty came from is a question this data leaves open. No timestamp records when each number entered a client's system, so a number that changed hands cannot be separated from one entered incorrectly or one that produced a greeting too thin to identify.

A phone number looks like one fact but holds two questions, whether the line still connects to the business named on the account, and whether anyone useful can be reached through it. The calls produced very different levels of evidence on each. Nonresponse is an outcome rather than a diagnosis. It records that communication failed without establishing why.

Verification answers the first question. When the supplied contact path does not produce a usable answer, the next step is the broader B2B skip tracing process, which follows the business entity, its principals, and the public record rather than treating one failed call as the end of the trail.

Methodology

This section describes the population, call protocol, coding framework, and limitations.

Population

JSD ran one verification attempt (a call, together with supporting review where the call alone left identity open) against each of 805 unique phone numbers attached to commercial accounts placed for collection, covering 805 distinct businesses. Calls ran between April 30 and August 26, 2026, on weekdays between 8am and 6pm Eastern. Because the file spans four time zones, the hour of the call relative to each business's own working day varies, and this study does not control for it.

Every number came from a client's own records and represents contact information supplied with the account. The numbers came from four client files. Every commercial account placed during the study window contributed its supplied number, and no accounts were excluded before calling, so this is a consecutive operational cohort rather than a sample drawn from a larger pool. The share contributed by each of the four clients is not reported here. Account size, account age, and industry went uncoded, so sector or client composition could weight the results in ways these figures would not reveal.

Numbers reached 48 jurisdictions. The largest shares were Florida at 16.1 percent, Texas at 11.6 percent, California at 8.3 percent, and New York at 4.6 percent, with no other jurisdiction above 4.1 percent and 4.2 percent of numbers carrying no jurisdiction value.

Exclusions

Two categories of call are excluded so that every observation represents one supplied number tested once. Some numbers received additional attempts during the same period, and the verification process itself surfaced additional numbers for some businesses, which were also called. Neither the repeat attempts nor the newly surfaced numbers are counted here, since the question is what one client-supplied phone number per business produces on a first verification attempt. No selection rule was needed, because each account arrived with a single contact number and every additional number tested was one the verification process itself turned up.

Verification is a standard step in JSD's placement workflow, run before an account is assigned to a collector. These calls were that step rather than an exercise built for this analysis, which is why every account placed during the window is present and none were excluded.

Coding

Each call carries two classifications. The endpoint records what answered, meaning a live person, a phone tree, a voicemail system, a carrier response, no answer, or a busy signal. The identity classification records what the call established about the number's connection to the account.

Where a greeting named a business under a trade name, franchise entity, or operating company, corroborating evidence decided whether that name belonged to the account relationship. Each number carries exactly one endpoint code and one identity code, both mutually exclusive. The evidence elements consulted during review, including greeting content and published web information matched against the account address, are non-exclusive supporting variables rather than categories.

Verification rested on the call evidence together with corroborating signals reviewed by hand, which is what allows a trade name, franchise entity, or operating company answering for a named account to be read as a match rather than a mismatch. Where a call arrived at a phone tree, the tree is the observation, and the question of who sits behind the menu stays open.

Identity results

The routable-but-unresolved and no-determination categories combine to 399 numbers, or 49.6 percent, for which the attempt did not establish whether the number belonged to the expected business.

Complete endpoint by identity cross-tabulation

Complete non-identification detail

The 441 numbers that neither verified the business nor put a person on the line, by endpoint:

Separately, 192 calls, or 23.9 percent of the file, never reached a business-controlled endpoint at all, meaning they rang out, returned busy, or drew a carrier response.

Carrier responses

Forty-six numbers drew a carrier response, and those split three ways.

The 39 numbers in the confirmed non-routable category are not all carrier responses. Thirty-eight arrived this way, and the thirty-ninth reached a voicemail system whose evidence established that the line no longer served the business.

A carrier response stating that the number was disconnected or no longer in service was treated as conclusive and coded non-routable. A response indicating network congestion, an invalid format, or another ambiguous failure was left as no determination, since those conditions do not by themselves establish that the number no longer exists. The underlying response codes do not separate the two groups beyond that threshold, so the distinction rests on review against that rule rather than on a fully mechanical classification. Coding all seven of the open cases as non-routable would move that category from 4.8 percent to 5.7 percent and no determination from 17.1 percent to 16.3 percent.

Line type

Line type came from carrier lookup data rather than from the call itself.

Mobile-classified numbers verified at 22.3 percent against a file average of 37.4 percent. With 256 numbers behind it, the gap is wide enough that sampling noise is an unlikely explanation, though the study establishes nothing about the cause. Client composition, industry, geography, account age, and how each number was originally captured all sit uncoded here, and any of them could carry the difference. Reading the result as a property of mobile lines would require reproducing it in a separate cohort. The smaller categories carry much less weight, and toll-free at 22 numbers would move several points on a handful of calls, so the ordering among the middle rows should be read loosely. Non-fixed VoIP covers internet-based numbers that carry no fixed service address, which includes many softphone and virtual office lines. Carrier lookup returned no usable line type for 197 numbers, or 24.5 percent of the file.

Limitations

The population consists entirely of commercial accounts that reached third-party collection, drawn from four client files, so this describes contact information at the point of escalation rather than commercial contact data generally. The concentration across those four clients is not reported.

The analysis does not estimate recoveries, resolution rates, right-party contact after follow-up, contact cost, payment outcomes, disputes, or the value of phone verification relative to other contact methods. It measures the condition of contact information at placement rather than collection effectiveness.

Because each supplied number was tested only once for this analysis, the reported verification rate should be read as the share resolved on a first attempt rather than the share that could ultimately be verified after repeated contact. A multi-attempt protocol would almost certainly produce a different figure.

Classifications came from operational review rather than independent blinded coding, and inter-reviewer agreement is not reported.

The dataset records no date for when each phone number entered the client's system, so it measures the condition of that data at placement while leaving the rate of decay unmeasured.

Client and debtor identifiers were excluded from the analysis. Figures are reported in aggregate only.