Short answer: Across 1.2 million real B2B website sessions measured over 90 days, visitor identification matched a company for about 47% of sessions and an actual individual for only 7%. Only 1 in 5 company matches was high confidence, and the match rate ranged from 13% to 65% depending on the site. There is no single "match rate." What you get depends on your traffic mix, your geography, and how much first-party signal you already hold.
Match rate is the most contested number in the visitor identification category. Vendors quote anything from 30% to 70%, usually with no methodology attached. So we measured it on our own network of live B2B websites and are publishing what the raw data shows, including the parts that are less flattering than a sales deck.
How we measured it
This study is built from first-party production data on the Abmatic AI platform. We looked at every real human website session over a rolling 90-day window ending July 6, 2026: 1,204,258 sessions across a portfolio of B2B websites. Simulated and internal test traffic was excluded. All figures below are aggregated and anonymized. No customer, domain, IP address, or individual is named or identifiable, and any breakdown cell with fewer than 1,000 sessions is suppressed.
We counted two distinct kinds of "match," because the industry blurs them on purpose:
- Company-level match: the session was resolved to a specific company (name and website), the way reverse IP lookup and firmographic enrichment work.
- Person-level match: the session was resolved to a specific individual (a linked contact), not just the organization they work for.
Those are not the same thing, and the gap between them is the single most important finding in this study.
Finding 1: About half of B2B sessions match to a company
Company-level match rate across all 1.2 million sessions was 46.8% by company name, and 51.2% when a resolvable company domain counts as a match. Call it roughly one in two. That also means more than half of B2B website traffic produced no company match at all, which is the honest counterweight to "identify your anonymous visitors" marketing. A large share of real traffic stays anonymous no matter which vendor you use, because the underlying signal simply is not there.
A blended 47% is a credible, defensible number for a mixed B2B audience. It is not the 70%+ some tools advertise, and any vendor quoting a single high number without a methodology is quoting a best case, not an average.
Finding 2: Person-level identification is far rarer, only 7%
Here is the number the category does not want in the headline. While 47% of sessions matched a company, only 7.0% resolved to an actual identified individual. A further 15.4% carried an inferred job title or role, which is a probabilistic persona, not a named person.
In plain terms: most "visitor identification" is company identification. You usually learn that someone from Acme Corp visited pricing, not who. That is still useful for account-based plays, but it is a different capability than "know the person," and buyers should price the two separately.
| What actually got matched | Share of sessions |
|---|---|
| Company (by domain) | 51% |
| Company (by name) | 47% |
| Inferred role or title | 15% |
| Named individual (linked contact) | 7% |
Finding 3: Only 1 in 5 matches is high confidence
A match is not a fact, it is a probability. When we graded every company match by the platform's confidence score, only 20.9% landed in the high-confidence band. Nearly 60% sat in the lowest bands. If you route sales outreach off every "identified" account without a confidence filter, most of what you act on is a coin-flip guess about who was really behind the visit.
The practical takeaway: judge a vendor on its high-confidence match rate, not its raw match rate. Raw rate is inflated by low-confidence guesses that look great in a dashboard and waste a rep's morning.
Skip the manual work
Abmatic AI runs targets, sequences, ads, meetings, and attribution autonomously. One platform replaces 9 tools.
See the demo →Finding 4: There is no single match rate, it ranges from 13% to 65%
Match rate is a property of your traffic, not just the vendor. Across the 8 sites in the sample with 1,000 or more sessions, company-level match rate ranged from 13.3% to 64.5%, with a median of 47.3%. Person-level ranged from 2.2% to 9.2%.
| Percentile across sites | Company match | Person match |
|---|---|---|
| Lowest site | 13.3% | 2.2% |
| 25th percentile | 33.3% | 3.2% |
| Median site | 47.3% | 6.4% |
| 75th percentile | 57.5% | 8.5% |
| Highest site | 64.5% | 9.2% |
A site running mostly direct and branded traffic from North American businesses will sit near the top. A site with heavy paid social, consumer overlap, or non-Western traffic will sit near the bottom. Same vendor, 5x difference in outcome. When someone promises you a match rate before seeing your traffic, they are guessing.
Finding 5: Geography collapses match rate
Where your visitors are changes everything. Match rate is highest in North America and lowest in privacy-strict and proxy-heavy regions. The table below is company-level match rate by visitor country, for countries with 1,000 or more sessions.
| Country | Company match rate |
|---|---|
| United States | 58.8% |
| France | 56.6% |
| Canada | 55.5% |
| India | 40.7% |
| United Kingdom | 34.1% |
| Japan | 34.0% |
| Spain | 23.6% |
| Australia | 22.9% |
| Germany | 13.9% |
| Singapore | 9.1% |
| Hong Kong | 5.1% |
| China | 4.0% |
Two forces are visible here. First, privacy regime and network behavior: German traffic matched at 14%, roughly a quarter of the US rate, consistent with stricter data practices and heavier VPN use. Second, traffic quality: origins dominated by datacenter and proxy traffic, such as China and Hong Kong, match near zero because the sessions are not real buyers sitting behind a corporate network. If your reporting counts those as "unmatched visitors you are missing," it is misreading bot noise as lost pipeline.
What this means if you are buying a visitor identification tool
The raw match rate on a vendor's homepage is close to meaningless without four qualifiers: is it company or person level, at what confidence, on what geography, and on what kind of traffic. A tool advertising 60%+ is almost certainly quoting company-level, low-confidence, US-heavy best case. Grade it on the number that changes your pipeline: high-confidence, actionable matches on your actual traffic.
And match rate is only half the question. Identifying an account is worthless if nothing happens next. The reason we can measure this at all is that identification on Abmatic AI is wired to action: a matched, high-confidence account can trigger visitor-level personalization, agentic outbound, and routing in the same platform, on first-party data you own rather than a third-party graph you rent. A 47% match that sits in a CSV loses to a 30% match that fires a play. Ask vendors what happens in the 60 seconds after a match, not just how often they match.
Frequently Asked Questions
What is a good visitor identification match rate in 2026?
For a mixed B2B audience, a blended company-level match rate around 40% to 55% is realistic and honest. Our measured figure across 1.2 million sessions was 47%. Rates above 60% usually reflect US-heavy, best-case traffic or count low-confidence guesses as matches.
Why is my match rate lower than the vendor promised?
Because match rate depends on your traffic, not just the vendor. Across sites we measured, it ranged from 13% to 65%. Heavy paid social, consumer overlap, non-North-American traffic, and mobile all pull the rate down. A vendor quoting a number before seeing your analytics is quoting an average that may not resemble your site.
Can visitor identification tell me the exact person who visited?
Rarely. In our data, only 7% of sessions resolved to a named individual, versus 47% to a company. Most "visitor identification" is company identification. Person-level resolution typically requires first-party signal you already hold, such as a prior form fill or CRM contact.
Does reverse IP lookup still work for this?
Partially, and less each year. Company-level resolution still leans on IP-to-company mapping, but datacenter IPs, VPNs, private relay, and mobile networks degrade it. See our breakdown of what reverse IP lookup can and cannot do in 2026. First-party identity is the more durable path.
How did you calculate these match rates?
From 1,204,258 real, non-simulated B2B website sessions on the Abmatic AI platform over a 90-day window ending July 6, 2026. Company match means the session resolved to a named company or domain; person match means it linked to a specific contact. All data is aggregated and anonymized, with cells under 1,000 sessions suppressed. No customer, domain, or individual is identifiable.
The honest summary
Half of B2B website traffic can be tied to a company, a small fraction to a person, and only when you also control what happens next does any of it turn into pipeline. Match rate is a starting condition, not a strategy. Anyone selling you a single big number is selling you the best case. The number that matters is how many high-confidence, actionable identities you get on your real traffic, and what your stack does with them in the next minute.
