Why Your Vendor's Match Rate Number Is Useless Without These Three Definitions

A website visitor identification match rate is the number of matched visitors divided by the population a vendor chooses to count. Without that denominator, the percentage cannot support a fair comparison. Buyers must separate three measurements, which are coverage, resolution level, and accuracy, to understand how much of their traffic a tool reaches, what it identifies, and whether the results are correct.

Why a single match rate percentage cannot answer the question it appears to answer

RB2B appears at several materially different published match rates, depending on who publishes the number and which denominator they use. The product does not need to change for those figures to differ. A tool that names the same 150 people can produce very different percentages simply by changing the population against which those matches are counted. Each calculation can be arithmetically correct. The missing information is what each percentage actually describes.

A July 2026 vendor survey found that, among 15 vendors with a published match rate, only three stated their denominator, one stated a sample size, and none published a confidence interval. Buyers get a figure that looks precise but does not tell them enough about what it covers or how much it could vary. A percentage alone cannot establish how the tool will perform on a particular website.

Coverage describes which visitors the tool has a chance to identify, while resolution level describes whether the result names an organization or an individual. Accuracy asks whether that result is correct. These are separate properties of the service being evaluated. Treating them as one score allows a strong result on one dimension to obscure a weakness on another.

How do website visitor identification tools calculate their match rate percentage?

Vendors calculate match rate against a population they define, often called “eligible” traffic. That population may exclude sessions without consent, non-US traffic, mobile and residential IPs, VPNs, privacy browsers, known crawler ranges, and internal visits. The exclusions change the fraction before a buyer sees it. Removing enough traffic can more than double the headline percentage without identifying a single additional person.

Some exclusions reflect genuine limits on where a tool works. Others remove traffic that would otherwise lower its reported yield. Either way, the distinction between an eligible-traffic rate and an all-traffic rate is essential to interpreting the result. A high rate on a restricted population says nothing by itself about how large that population is.

Leadpipe’s match rate glossary says that some vendors count sessions rather than unique visitors, so a person who returns three times is counted in the session denominator three times. Sessions and people are not interchangeable measurement units. Whether the resulting percentage rises depends on how matches are counted in the numerator as well. Without that corresponding definition, a session-based rate cannot be compared directly with a visitor-based rate.

The practical test is a trial on the buyer’s own raw, unfiltered traffic, reporting matched visitors as a share of total unique visitors. An eligible-traffic figure can help show the effect of exclusions if it is reported separately. The gap between the two exposes how much the exclusions change the headline number. However, normalizing the denominator only establishes a common starting point.

Variable one — coverage measures what share of actual visitors the tool reaches at all

Coverage is the share of real human sessions the tool gets a fair chance to match. Remote work, VPNs, mobile networks, and privacy browsers can all restrict that opportunity. An identity graph cannot resolve a signal it does not receive. A visit may go unmatched because the infrastructure fails to provide a signal, even when the matching process works well.

Reverse IP lookup makes that distinction concrete. As the foundation of company-level identification, it depends on a visitor arriving from a static corporate IP range registered to an employer. A knowledge worker using a home Comcast connection is invisible to that method. Public Wi-Fi, mobile networks, and proxy servers can leave the same gaps. The signal is absent.

Independent tests in 2026 show that realistic company-level match rates are 30–65% of US B2B traffic. Remote work, shared office spaces, and VPNs help explain why those results fall below vendor claims by moving employees off corporate IP ranges. That range describes a particular traffic population, which is precisely the qualification a headline percentage often omits. A buyer with a different traffic mix needs results from that mix.

Bot and crawler traffic complicates the accounting further. Including crawlers in the denominator lowers the reported match rate, while filtering them out raises it without improving human-visitor coverage. Separating human sessions from automated traffic makes the coverage measurement more interpretable. The raw-traffic trial and the human-coverage measure answer related but distinct questions about total yield and the opportunity to identify real visitors.

Variable two — resolution level determines whether “matched” means a company or a named person

Company identification and person identification are different products. Company-level identification resolves the organization behind a visit, primarily through reverse IP lookup against an IP-to-company database. Person-level identification names the individual, requiring identity-graph or first-party signals layered on top of the IP match. One result does not establish the other. Leadpipe’s definition accordingly separates company match rate from person match rate.

Vibe Prospecting’s 2026 guide draws a useful distinction between contact enrichment and visitor identification. Enrichment fills in fields on a record already held, while identification builds a record from an anonymous signal. Person-level resolution is the harder, lower-yield version of that second task. Independent tests in 2026 show that realistic person-level rates are 5–20% of B2B traffic.

Geography further constrains person-level results because identity-graph data is US-sourced, and identification works overwhelmingly on US-based B2B traffic. A high person-level rate on “US B2B traffic” and a lower rate on “all traffic” may describe identical underlying performance. The apparent difference can come entirely from scope. This is why resolution and denominator must be disclosed together.

Combining resolution levels creates another comparison problem. RB2B’s support documentation lists contact-level and company-level identification for US traffic separately, then describes an overall figure that appears to sum them. Those populations can overlap. Adding the rates without accounting for that overlap does not establish the share of distinct visitors identified. For sales, the distinction is practical too, because knowing that Acme Corp visited a pricing page does not reveal whether the visitor was a junior researcher or the VP of Revenue Operations.

Variable three — accuracy measures how many returned matches are actually correct

Accuracy measures the correctness of the matches returned. A tool can raise its reported match rate by lowering the confidence threshold at which it returns an identity. More results then come with more false positives. Leadpipe’s glossary describes this trade-off directly, contrasting permissive thresholds with stricter thresholds that return fewer, more accurate matches.

Deterministic matching ties a visitor to a confirmed signal, such as a form fill, a matched email, or a verified cookie. Probabilistic matching instead correlates IP ranges, device fingerprints, and behavioral patterns to infer identity. With probabilistic matching, headline rates can be higher while accuracy is lower. Buyers need to know which method and confidence threshold produced the published number.

Accuracy also changes as contact data ages. B2B contact data goes stale materially every year, so a measurement taken at launch or signup may overstate the quality available later in a contract. Initial performance is not a permanent guarantee. The practical test is to sample returned matches and check email deliverability against the identified company domains. A high bounce rate is a warning that the confidence threshold may be too permissive, although deliverability alone does not prove that the named person made the visit.

How inflated match rates corrupt paid-media attribution downstream

Paid-media attribution depends on the identity layer correctly resolving who visited. If an inferred identity is wrong, assigning a campaign source to that identity does not repair the error. It moves the error downstream into another report. The promise of connecting a LinkedIn or Google impression to a named company’s later visit therefore depends on the underlying match being genuine. A probabilistic guess can receive conversion credit without establishing a causal relationship to the ad spend.

A narrowed denominator can make identification appear more comprehensive than it is, while a permissive confidence threshold can introduce incorrect identities into campaign reporting. The first distorts the apparent scope of measurement. The second can distort which accounts receive paid-media credit. Treating either headline rate as sufficient evidence leaves return-on-ad-spend calculations resting on an unexamined identity layer.

Accounts credited to a paid campaign may include visitors who would have found the site organically, so paid spend receives credit for activity it did not cause. That attribution can make a campaign appear more productive than the underlying evidence supports. Budget decisions then inherit the identification assumptions. A match rate cannot validate those decisions unless its population, resolution level, and confidence standard are understood.

Five questions that force a vendor to show you their denominator

Before signing, buyers should ask five questions to help them compare the proposed service with other options. The first is about the denominator. Does the published percentage use all sessions, unique visitors, or only traffic that passes consent and geography filters? The answer should name both the counting unit and the exclusions. “Eligible traffic” alone is not specific enough to establish what the rate measures.

The second question asks whether the figure measures company matches or person matches. The vendor should state each rate separately. If an overall identification figure is also provided, the explanation should account for overlap between the two populations. This establishes what the buyer will actually receive. A named organization and a named individual support different sales actions.

Third, buyers should ask what confidence threshold triggers a match and whether the published rate uses that same threshold. Evaluation and production need a common standard. A lower threshold during a demonstration can produce a higher rate than the buyer experiences after signing if production uses a stricter rule. The answer should make that trade-off explicit. Otherwise, the trial may measure a different balance of yield and accuracy from the purchased service.

Fourth, buyers should ask whether they can test the returned records for accuracy, including email deliverability against the identified company domains. A sample with a high bounce rate warrants scrutiny of the threshold behind it. Receiving more records is not the same as receiving more usable identities.

Fifth, buyers should ask whether the vendor will run a trial on the buyer’s own unfiltered traffic and report matched visitors as a share of total unique visitors. An eligible-traffic result can be reported separately for context.