A visitor identification tool is only useful if the companies and people it names are the right ones. Before matches reach sales, a RevOps or marketing ops team needs a repeatable way to prove that, using visits it can already identify.
This works when the site gets enough form submissions or logins to supply known visits, and when the tool lets you export its raw match data. Known visits come from people who convert and accept cookies, so correctness measured on them is a best case for anonymous traffic.
What Does Visitor Identification Accuracy Mean?
Visitor identification accuracy is two numbers, not one. Coverage is the share of engaged sessions a tool such as Snitcher, Leadinfo or RB2B names, and correctness is the share of those names that prove right when checked against visits you can already identify.
Most of what is published about website visitor identification is a match rate, and a match rate is a coverage figure. It says how many visitors a tool put a name to. It says nothing about whether those names were right.
Match rates also use different denominators. Knock2, a competing vendor, points out that published RB2B match rates range from 5 to 20% up to more than 40%, a spread of nearly ten times, because each source divides by a different count of visitors. Knock2's fix is to divide identified visitors by engaged sessions on your own traffic, meaning visits of 10 seconds or more, or two or more pageviews, as Google Analytics defines them.
A third number belongs beside the first two. Leadpipe's measurement framework, another vendor's, marks every identity correct, incorrect or unknown, and reports the unknown count next to any correctness rate. A correctness figure without the unknown count hides how much of the sample could not be checked at all.
So a validation test produces three outputs. Coverage on engaged sessions, correctness on known visits, and the count of matches you could not confirm either way. A vendor's published match rate is none of these.
What Are Known Visits?
A known visit is a session whose visitor you can name from their own action, such as a HubSpot form submission that ties earlier page views to a contact through the tracking cookie, so it can serve as ground truth for checking what an identification tool reported.
HubSpot's tracking code records visitors anonymously before they become contacts. When a visitor submits a form, HubSpot associates their previous page views with the new contact through the tracking cookie. That gives you a timestamped list of pages viewed by a person whose email, and therefore whose company domain, you know. The tracking code is available on every HubSpot product and plan.
Other sources of known visits:
- Tracked email clicks. A click in a HubSpot email associates the cookie with an existing contact, but only once.
- Logins. A product or portal login names the person directly.
- Snitcher user identification. Snitcher identifies users when they submit a form, log in, click a tracked email link or are passed through
Snitcher.identify(), and merges their profiles automatically.
Some known visits disappear before the test starts. A visitor who deletes cookies is treated by HubSpot as a new visitor. If a contact does not accept cookies, their page views do not appear on the contact record. HubSpot also does not track unique visitors, so session counts have to come from your analytics tool.
The rule that matters most is where the identity comes from. A known visit is named by the visitor's own action. A company that an identification tool named, even one that looks right, is never a known visit, because using one tool's matches to grade another only measures how often they agree.
Why Do Visitor Identification Matches Go Wrong?
About a quarter of US paid workdays are now worked from home, according to the Survey of Working Arrangements and Attitudes, and a home connection carries a consumer ISP address that IP-based tools such as Snitcher and Leadinfo cannot tie to an employer.
26%
Share of paid US workdays worked from home in July 2026. Among full-time employees, 12% were fully remote and 26% hybrid.
July 2026 · US residents aged 20 to 64 who earned at least $10,000 in a prior year · Source
Company-level identification works by matching a visitor's IP address to a company that owns or uses it. That works well on office networks and breaks down in four common situations.
- Shared consumer ISPs and mobile carriers. Snitcher's documentation says consumer ISPs are shared by millions of households and mobile carrier IPs give limited accuracy. Leadinfo notes that consumer ISPs assign dynamic IPs that are hard to link to a company.
- Home working. A remote employee on home broadband looks like a household, not a business. Leadinfo says home workers are harder to trace unless they use a company VPN or fixed IP.
- VPNs and proxies. Consumer VPN services mask location. Corporate VPNs that route through company infrastructure are the exception, and Snitcher rates accuracy on them as high.
- Apple iCloud Private Relay. Private Relay sends Safari traffic through two relays, and the second, Apple-operated relay generates a temporary IP address that connects the user to the site. Only the network provider and the first relay see the real IP.
Tools handle doubt differently, and that shapes what a test finds. RB2B abandons identification when a visitor uses a VPN or proxy, because it cannot confirm the visitor is in the United States. Leadinfo says that when in doubt, a visitor is not shown as identified. Both choices trade coverage for correctness. You will see fewer matches, but fewer guesses among them.
That is why the mismatch investigation later in this playbook asks for the cause of each wrong match. Most causes are network causes, and they point to different fixes.
How To Validate Visitor Identification Accuracy Step-By-Step
To validate visitor identification accuracy, run the tool on the same pages over a fixed window, then check every match it made against HubSpot visits you can already name, because two tools agreeing proves little when one benchmark found only a 4% overlap between the companies RB2B and Clearbit Reveal identified.
That benchmark came from MadKudu on its own website, with no sample size or period published, so treat it as an illustration of how little tools agree rather than a market rate. The point holds either way. Agreement between providers is not ground truth, which Leadpipe's framework also states.
- Define coverage, correctness and the unknown count, with engaged sessions as the denominator
- Export known visits from HubSpot for the test window, with contact email domain and visit timestamps
- Run each tool on the same pages, from the same start date, for the same period, and keep the raw export
- Join each match to known visits by timestamp and page, then compare matched domain with the contact's email domain
- For person-level matches, compare the named person with the HubSpot contact for that visit
- Mark each match correct, incorrect or unknown, and report the unknown count
- Calculate coverage, correctness and cost per correct identification
- Send every incorrect match to the mismatch investigation
Step three follows Snitcher's own guidance for comparing providers. Install every script on the same date, track the same pages, run trials for the same duration and insist on full access to the raw trial data. Snitcher's help centre warns against any vendor that withholds the raw data set and offers to run the competitor analysis itself. A tool whose raw matches you cannot export cannot be validated, only demonstrated.
Knock2 suggests a 14-day concurrent test on your own traffic. Whatever window you pick, keep it identical across tools so seasonality and campaigns affect each one equally.
The join in step four is where most false errors appear. A match that lands on a different session from the known visit is a join problem. Check timestamps before you judge the tool.
Correctness is the share of checked matches marked correct. Coverage is matches divided by engaged sessions. Cost per correct identification is the plan cost, or resolutions spent, divided by the matches marked correct.
For what independent studies have found on known accounts, read our research on visitor match accuracy on known accounts. This playbook covers the operating side, running the test on your own site.
How To Build The Visitor Identification Workflow In HubSpot And Airtable
A workable visitor identification workflow uses HubSpot for known visits, the identification tool for matches and an Airtable base as the log, where each match becomes one record with its timestamp, page, matched company, known contact and verdict.
Airtable suits the log because each match is a row a reviewer can filter, label and count. Give the table these fields:
- Tool
- Visit timestamp and page
- Matched company or person
- Known HubSpot contact and email domain
- Verdict (correct, incorrect, unknown)
- Mismatch cause
- Confidence label
Matches can reach Airtable automatically through the "When webhook received" automation trigger. It creates a unique URL that runs an automation when a POST request arrives. Payloads are capped at 100 KB per request and five requests per second, and every run counts toward the workspace's automation run limit. Airtable does not support signature verification for this trigger, so keep the URL private. Tools without a direct webhook can pass matches through a connector. Leadinfo, for example, connects to Zapier and Make as well as HubSpot.
The log has real ceilings. Airtable's Team plan allows 50,000 records per base, 25,000 automation runs per month per workspace and 100,000 API calls per workspace per month. When a base exceeds its record limit, Airtable blocks new records until an upgrade, so a busy site should log a sample rather than every match.
Permissions and limits for a validation run
| Part | Who can set it up | What the test consumes |
|---|---|---|
| HubSpot tracking and form contacts | Tracking code on every page | Included in every HubSpot plan |
| HubSpot buyer intent | Buyer Intent permission, assigned by a Super Admin | Credits beyond 20 companies per view |
| Adding companies from buyer intent | Data Enrichment permission | Credits to add and track companies |
| Airtable automations | Base Owners and Creators | Automation runs toward the plan limit |
| Airtable validation log | Base Owners and Creators | Records per base |
| Leadinfo to HubSpot | An automation slot, manual on Starter | Automations per plan |
| RB2B matches | US traffic for person-level ID | Monthly resolutions, then overage |
| Snitcher matches | Script on the same pages | Unique companies identified |
The validation run is not free. HubSpot credits, Airtable automation runs, RB2B resolutions and Snitcher's identified-company count all rise with the size of the test, so budget for it before the window opens.
Airtable pricing
| Plan | Price | Billing | Notes |
|---|---|---|---|
| Team |
|
| per seat |
| Business |
|
| per seat |
| Enterprise Scale | Custom pricing | per seat | |
| Free | Custom pricing | per seat |
How To Run A Mismatch Investigation
A mismatch investigation starts by lining up timestamps, because a match that falls on a different visit from the known one is a join error, not a tool error, and only after that does the reviewer compare domains and look for a cause such as a shared ISP or a VPN.
Work through every incorrect or doubtful row in the same order:
- Timestamp. Does the tool's visit fall on the same session as the known visit? If not, fix the join and re-check.
- Domain. Does the matched company's domain match the contact's email domain? If not, look for a cause.
- Cause. Record the cause in the log, then take the action that cause calls for.
Mismatch causes and actions
| Cause | Evidence in the record | Action |
|---|---|---|
| Join error | Match falls on another session | Re-join on timestamp and page |
| Shared ISP or mobile carrier | Company is an ISP, location approximate | Reject and exclude |
| VPN or iCloud Private Relay | Company is a VPN provider, or visit missing | Mark unknown, check Safari share |
| Home worker | Right person, wrong or no company | Mark company match incorrect |
| Parent or subsidiary | Domain differs, company related | Mark correct, note the relation |
| Stale firmographics | Right company, outdated details | Send raw rows to the vendor |
| Data-centre or bot traffic | Hosting company, no engagement | Reject and exclude |
| Wrong known record | Personal or shared email domain | Check the form source first |
There are three actions to choose from. Exclude a recurring false company, for instance through HubSpot buyer intent exclusions, which accept up to 100 companies. Ask the vendor about it with the raw rows attached. Or relabel the match's confidence in the log.
Never assume the known record is the wrong one without checking where it came from. A form filled in with a personal email address or by an agency on a client's behalf can make a correct match look wrong.
Patterns matter more than single rows. A high share of VPN or Private Relay causes points to an audience that no IP-based tool will cover well. A cluster of ISP names points to filters the team should add.
How To Assign Confidence Labels To Visitor Matches
Because Snitcher, Leadinfo, RB2B and HubSpot buyer intent publish no per-match confidence score, a team has to label each match itself, and a person-level match backed only by a professional profile is weak evidence, since the profile confirms an employer but not the visit.
Labels should come from the evidence each match carries in the Airtable log, not from how sure a tool sounds. Four labels are enough.
Confidence labels for visitor matches
| Label | Evidence | What sales may do |
|---|---|---|
| Confirmed | Matches a known visit | Act on it |
| Likely | Office network or corporate VPN, repeat visits, fits ICP | Research, then reach out |
| Weak | Single visit, consumer ISP, or profile-only person match | Nurture only |
| Rejected | Failed the mismatch investigation | Nothing, exclude it |
The person-level case needs care. RB2B enriches verified profiles from professional sources such as LinkedIn. Leadpipe's framework notes that a professional profile can confirm an employer or a title, but not that the person visited. A named person with no other supporting evidence therefore stays weak until a known visit or repeat activity backs it up.
Company-level and person-level matches rest on different evidence and fail in different ways. For how the two levels differ in coverage and confidence, see our guide to company-level vs person-level visitor identification.
Apply labels in the log before any match is routed. Sales then works confirmed and likely matches first, and the team can count how many of each label a tool produces per month.
Validating Snitcher Matches
Snitcher names companies from a database of business IP ranges and shows unmatched visits as Anonymous rather than guessing, so its matches can be checked against the visitors Snitcher itself identifies through forms, logins and tracked email links.
Snitcher: User identification In the documentation
Snitcher's documentation describes identifying users when they submit a form, log in, click a tracked email link or are passed through Snitcher.identify(), with profiles merged automatically. That gives a second known-visit source inside the tool.
How it matches: Snitcher identifies companies from a proprietary, continuously updated database of business IP ranges. It works at company level, not individual level, and its documentation publishes no accuracy percentage.
Known-visit check: Compare each identified user's email domain with the company Snitcher matched for the same session, and cross-check against HubSpot form contacts. Snitcher's own guidance for comparing providers also lists accuracy as how many conversions, such as form fills, were correctly attributed.
Where matches go wrong: Snitcher rates accuracy as very high on office networks with dedicated IP ranges and high on corporate VPNs, but limited on mobile carrier IPs. Consumer ISPs and VPN services are the weak spots.
Built-in safeguards: A visitor Snitcher cannot identify appears as Anonymous with an approximate IP location, so unmatched traffic is visible rather than filled with a guess.
What drives the cost of a test: The Platform plan is priced by unique companies identified each month, in tiers. A validation run costs more as it identifies more companies, not as you add seats or sessions.
Snitcher Platform pricing
| Plan | Price | Billing | Notes |
|---|---|---|---|
| Platform 0-50 company IDs |
|
| |
| Platform 51-100 company IDs |
|
| |
| Platform 101-250 company IDs |
|
| |
| Platform 251-500 company IDs |
|
| |
| Platform 501-750 company IDs |
|
| |
| Platform 751-1,000 company IDs |
|
| |
| Platform 1,001-2,000 company IDs |
|
| |
| Platform 2,001-3,000 company IDs |
|
| |
| Platform 3,001-4,000 company IDs |
|
| |
| Platform 4,001-5,000 company IDs |
|
| |
| Platform Custom | Custom pricing |
Validating Leadinfo Matches
Leadinfo hides a visit when its sources disagree and keeps no IP addresses, so a reviewer validates it by comparing each named company with HubSpot known visits on timestamp and domain rather than by inspecting the raw IP.
Leadinfo: ASN mapping Vendor states this
Leadinfo states that it uses ASN mapping to find companies with multiple IP ranges or cloud infrastructure, and that every identification is checked against multiple data sources before it is shown.
How it matches: Leadinfo matches a visitor's IP against its databases and uses ASN mapping for companies with several IP ranges. Leadinfo claims a 35 to 40% identification rate by its own measure, lower for visitors outside Europe. That is a coverage figure, so measure coverage on your own engaged sessions instead.
Known-visit check: Leadinfo connects to more than 70 tools, HubSpot among them. Push its matches into HubSpot or the Airtable log and compare each company with known visits on timestamp and email domain.
Where matches go wrong: Leadinfo's help centre says visits from home networks may not be recognised as a company, mobile connections use hard-to-link dynamic IPs, and only LinkedIn ad clicks that land on the site are recognised. Insecure HTTP pages can break the tracking script.
Built-in safeguards: When in doubt, Leadinfo does not show a visitor as identified. It does not save IP addresses, and companies can block their data from being shown, so a reviewer works from timestamp and company rather than the raw IP.
Access and permissions: Plans differ in seats and automations. Starter includes 3 user seats and 3 automations, Scale 10 and 10, and Pro unlimited. CRM sending is manual on Starter, so an automatic push to HubSpot needs an automation slot.
What drives the cost of a test: Plans are tiered by companies identified per month. Paid add-ons such as Autopilot and Leadbot Business start from EUR 59 per seat, and Liquid content from EUR 39 per month.
Leadinfo pricing
| Plan | Price | Billing | Notes |
|---|---|---|---|
| Starter |
|
| |
| Scale |
|
| |
| Pro |
|
|
Validating RB2B Matches
RB2B identifies people and companies through independent processes, with person-level matches limited to visitors it can place in the United States and company-level matches available globally, so each level needs its own correctness check against HubSpot known visits.
RB2B: Visit activity log In the documentation
Each RB2B visitor profile has an Activity section listing the date and time of every visit, which lets a reviewer line up an RB2B match with a known HubSpot visit.
How it matches: RB2B cross-references first- and third-party cookies, device identifiers and IP address analysis, focused on US traffic, and enriches verified profiles from professional sources such as LinkedIn. Company-level profiles carry no personal identifiers. RB2B claims it identifies about 70 to 80% of US traffic at contact level, on its own denominator.
Known-visit check: Use the profile's Activity list to find the visit, then compare the named person with the HubSpot contact for the same timestamp. Run the company-level check as a separate test, on all traffic with no US restriction.
Where matches go wrong: Visitors on a VPN or proxy are dropped, so coverage falls on VPN-heavy audiences. A person match supported only by a profile confirms employer, not the visit.
Built-in safeguards: RB2B abandons identification when it cannot confirm a visitor is in the United States, rather than guessing.
What drives the cost of a test: RB2B bills by monthly resolutions. Overage is 45 cents per resolution on Starter and 25 cents on Pro and Pro+, so a long test on a busy site can run into overage. Company-level ID is global on every plan, while contact-level ID is US only.
RB2B pricing
| Plan | Price | Billing | Notes |
|---|---|---|---|
| Free | $0 | per month | |
| Starter | $79 | per month | |
| Pro | $149 | per month | |
| Pro+ | $199 | per month |
Validating HubSpot Buyer Intent Matches
HubSpot buyer intent matches anonymous visits to companies by IP address, so its results have to be checked against HubSpot contacts identified through form submissions, never against other buyer intent records.
HubSpot: Buyer intent In the documentation
HubSpot's knowledge base describes buyer intent as connecting anonymous web visitors to known companies' IP addresses, with filters for the In CRM status and visitor country, and a Buyer-Intent record source on companies it adds.
How it matches: HubSpot plays two roles in this test. Its form-tracked contacts are the known visits, and its buyer intent tool is itself an IP-based company identifier. HubSpot states no accuracy rate for it.
Known-visit check: Use the In CRM filter to find buyer intent companies that already have contacts, then check those contacts' form-tracked page views for the same pages and dates. Companies added from buyer intent show a Record source of Buyer-Intent, which keeps them apart from known records.
Where matches go wrong: Because buyer intent relies on IP addresses, it meets the same limits as other IP-based tools. Home workers, consumer VPNs and Private Relay traffic are the likely gaps, and those are the rows to investigate first.
Built-in safeguards: Filters cover the last 90 days in UTC, visitor country, page paths, In CRM or Not in CRM, lifecycle stage and owner. Up to 100 companies can be excluded, which is where recurring false matches belong.
Access and permissions: Buyer intent is available on Marketing, Sales, Service, Data and Content Hub Starter, Professional or Enterprise. Configuring it needs Buyer Intent permissions assigned by a Super Admin, and adding companies needs Data Enrichment permissions.
What drives the cost of a test: Viewing more than 20 companies per view, adding companies and tracking their intent consume HubSpot credits. Marketing Hub includes 500 credits on Starter, 3,000 on Professional and 5,000 on Enterprise. Professional also carries a one-time onboarding fee of $3,000 and Enterprise $7,000.
HubSpot pricing
| Plan | Price | Billing | Notes |
|---|---|---|---|
| Free | $0 | per month | per user |
| Starter |
|
| per seat |
| Enterprise | $3,600 | per month |
If the test shows HubSpot buyer intent falls short, compare the alternatives in our guide to visitor identification tools for HubSpot.
What Success Measures Should You Track For Visitor Identification?
The measure that decides whether a visitor identification tool pays its way is cost per correct identification, the plan cost or resolutions spent divided by the matches that proved right on known visits, reported with the count of matches that could not be checked.
Success measures for visitor identification
| Measure | How to calculate | Decision it feeds |
|---|---|---|
| Coverage | Matches divided by engaged sessions | Whether the tool sees enough traffic |
| Correctness | Correct divided by checked matches | Whether sales can trust matches |
| Unknown count | Matches you could not check | How far correctness can be trusted |
| ICP identification rate | ICP matches divided by engaged sessions | Whether matches are the right accounts |
| Cost per correct identification | Plan cost or resolutions divided by correct matches | Whether the tool pays its way |
| Sales acceptance | Matches sales accepts divided by routed | Whether routing rules work |
| Time to CRM | Visit time to record creation | Whether alerts arrive in time |
Snitcher's help centre lists ICP identification rate and price per identification among the measures for comparing providers. Leadpipe's framework adds sales acceptance and delivery, which test whether a correct match is also a useful one.
Report ranges and counts, not a single percentage. Leadpipe notes that a handful of successful matches can demonstrate a workflow but cannot establish a narrow error rate. A test with few known visits should say so beside its result.
Re-check after any site redesign, a shift in traffic mix such as a new paid campaign or region, or a plan change at the vendor. Outside those events, pick a fixed cadence the team can sustain, and run the same test each time so results stay comparable.
A correct match still tells you who visited, not why. Keep in mind that a website visit does not always mean buying intent before routing every confirmed company to sales.
FAQ
Can you trust a vendor's published match rate?
Not as an accuracy figure. A vendor's published match rate cannot be compared with another vendor's, because each divides identifications by a different count of visitors. It measures coverage, not correctness. The only comparable figure is the one you measure on your own engaged sessions, checked against known visits from your own CRM.
How many known visits do you need for a validation test?
Enough to report a range rather than a single number. A handful of correct matches can show the workflow runs, but it cannot establish a narrow error rate. Run the test long enough to collect a meaningful set of form-tracked visits, and always publish the count of checked and unknown matches beside the correctness figure.
Does cookie consent affect known visits?
Yes. HubSpot ties earlier page views to a contact through its tracking cookie, so if a contact does not accept cookies, their page views do not appear on the contact record. Visitors who delete cookies are treated as new. Known visits therefore come mostly from people who accept cookies, which makes correctness on them a best case.
Sources
Sources
- WFH Research Updates, September 2026 · wfhresearch.com
- About iCloud Private Relay · support.apple.com
- An Unbiased Review and Benchmark of RB2B · madkudu.com
- leadinfo.com
- help.snitcher.com
- docs.snitcher.com
- snitcher.com
- leadinfo.com
- airtable.com



