When a website visitor identification vendor tells you they identify "90% of your traffic," ask one question before you sign anything: 90% of what? Most accuracy claims in this category have no defined denominator, blend company-level and person-level numbers into one headline figure, and can't be reproduced against your own traffic. If a vendor can't answer "measured against what, and for which traffic profile," treat the number as marketing copy, not a spec.

This is a buyer-side skepticism problem the category has earned. Here's how to vet a claim before you buy, a checklist for spotting the red flags, and a two-week test you can run yourself this week.

What Does "95% Accurate" Actually Mean?

An identification rate is a fraction. The numerator is usually clear: sessions the vendor successfully matched to a company or a person. The denominator is where vendors get vague, and the denominator changes the number by 3 to 5x depending on which one a vendor picks.

Common denominators, from most to least favorable to the vendor:

• All traffic, including bot hits, single-second bounces, and international visitors the vendor never had a shot at identifying. This produces the lowest, most "honest" looking numbers, so it's rarely what inflated claims are measured against.
• All sessions, which filters bots but still counts visitors who bounced in under a second, a group no identification method can realistically resolve.
• Engaged sessions, typically defined as a visit lasting 10 seconds or longer, or two or more pageviews. This is the denominator that best reflects a visitor who actually looked at your site, and it's the one a defensible claim should disclose.

A vendor quoting a rate without naming the denominator is choosing not to tell you which of these they used. That's the first red flag, not a technicality.

Company-Level and Person-Level Are Not the Same Number

The second most common way accuracy claims mislead is collapsing two very different measurements into one headline stat. Company-level identification (matching a visiting IP or device to a business) and person-level identification (resolving an actual name, title, and email) are separate problems with separate ceilings. Company-level rates in the 80 to 95% range are achievable for B2B-heavy traffic. Person-level rates in that range are not, no matter what a landing page implies.

If you're deciding which type of match actually serves your workflow, the tradeoffs are worth a closer look in our breakdown of person-level vs. company-level identification. For this post, the point is narrower: any claim that doesn't specify which type it's measuring should be read as a company-level number dressed up to look like a person-level one.

The Red-Flag Checklist for Vendor Accuracy Claims

Use this severity scale when a vendor hands you a number. The more red a claim scores, the more proof you should demand before it factors into your decision.

• 🔴 Very High - A single blended percentage with no denominator, no traffic-profile caveat, and no distinction between company- and person-level matching.
• 🟠 High - A denominator is named, but it's "all traffic" rather than engaged sessions, inflating the number against a base the vendor never had to identify.
• 🟡 Medium - Denominator and match type are both disclosed, but the number is a global average with no note on geography or industry, both of which move identification rates materially.
• 🟢 Low-Medium - Full methodology is disclosed and the vendor offers to validate the number against a sample of your own traffic before contract signature.
• ⚪ Low - The vendor publishes the methodology publicly, invites a side-by-side test, and the number holds up when you run it yourself.

How to Run Your Own Two-Week Identification Bake-Off

You don't need a data team to verify a vendor's claim. You need two weeks, your own traffic, and a spreadsheet.

• Step 1: Instrument the same pages with two tools. Run the incumbent or top-choice vendor alongside a second option on your highest-traffic pages for the full two weeks. Don't sample, run it concurrently, so both tools see the same visitors.
• Step 2: Pull your engaged-session count from analytics. Use your own analytics platform's definition of an engaged session (10+ seconds or 2+ pageviews) as your denominator, not the vendor's dashboard number.
• Step 3: Export identified records from both tools and de-duplicate against your CRM. Count how many identified companies or contacts are already in your CRM as noise (existing customers, your own team, competitors doing research) versus net-new ICP accounts.
• Step 4: Score match quality, not just match volume. A 40% match rate full of accounts outside your ICP is worse than a 20% rate that's clean. Sort identified companies by fit before you sort by count.
• Step 5: Calculate your own rate. Divide net-new, ICP-fit identified sessions by total engaged sessions. That's your real number, and it's the only one that should go into a vendor decision.

What a Defensible Accuracy Claim Looks Like

This is the standard we hold ourselves to. Knock2 publishes two identification rates, both measured against engaged sessions: 93%* account and company-level identification, and 62%* person-level identification (name, email, title) for US traffic. Both numbers carry the same footnote: identification rates are measured against engaged sessions, and results vary by traffic profile, geography, and industry.

Notice what that claim does that a vague one doesn't. It names the denominator (engaged sessions). It separates company-level from person-level rather than blending them. And it tells you up front that your number will move based on your traffic, not stay pinned to the headline figure regardless of who's visiting your site. That's the bar to hold any vendor to, including us.

On the mechanism: person-level identification works by matching visitors against an identity graph built from a consent-based publisher network, not by claiming to unmask anonymous individuals from thin air. If a vendor won't explain even that much about how their match happens, that's worth asking about directly.

Why Identification Rates Vary by Traffic Profile

The reason "your mileage may vary" isn't a hedge, it's math. A US-based B2B SaaS company with a heavily gated, high-intent traffic mix (think a dev-tools or fintech product with a technical buyer) will consistently see higher person-level match rates than a company with broad international or consumer-adjacent traffic. We see this across our own customer base: teams selling into narrow, US-concentrated technical buyer personas post materially higher person-level rates than teams with global, top-of-funnel content traffic.

This is exactly why running the two-week test above against your own traffic matters more than any published benchmark, including ours. Use published rates to sanity-check a vendor's methodology. Use your own bake-off to make the buying decision.

Once you've got identification working, the next problem is turning matched visitors into pipeline before the moment passes. Our six plays for turning anonymous traffic into pipeline and our breakdown of how much of your website traffic is actually your ICP are the natural next reads. If you're still mapping the broader category, our comparison of engagement tracking tools is a useful companion to this checklist.

FAQ

What is a good website visitor identification accuracy rate?

For company-level identification against engaged sessions, 80 to 95% is achievable for B2B-heavy US traffic. For person-level identification, rates in the 50 to 65% range against engaged sessions, US traffic only, are strong. Anything person-level above that, without a clearly disclosed methodology, deserves scrutiny.

Why do identification rates differ between vendors?

Different vendors use different denominators (all traffic vs. sessions vs. engaged sessions), different identity graph coverage, and different traffic profiles in their published examples. Two vendors quoting different numbers may both be telling the truth about different measurements.

Should I trust a vendor's published accuracy number at all?

Treat it as a starting point, not a decision input. A published number tells you whether a vendor is willing to disclose methodology. Your buying decision should be based on a bake-off against your own traffic, not the vendor's marketing page.

How long should a visitor identification vendor evaluation take?

Two weeks of concurrent testing is usually enough to get a statistically meaningful read for most B2B sites, longer if your traffic volume is low or highly seasonal.

Does a higher identification rate always mean a better tool?

No. A high rate full of out-of-ICP accounts and consumer traffic is worse than a lower rate that's clean. Match quality, measured by how many identified accounts are actually your ICP, matters more than raw match volume.

Ready to see your own numbers instead of a vendor's headline claim? Book a Knock2 demo and run the bake-off against your own traffic before you decide.