What is a realistic match rate for person-level website visitor identification?
Visitor identification match rates depend on who visits your site. On business traffic, person-level identification tools commonly resolve 5–20% of visitors, and B2B guidance often suggests planning around 10–15%.
On US consumer traffic, rates run much higher. Untitled identifies 40–60% of US consumer traffic at the person level. The exact rate depends on site type, traffic volume, and how deeply visitors engage.
For context, a website on its own, relying on forms and logins, typically identifies only 1–2% of its visitors.
Keep two numbers separate from the start. Match rate tells you how many visitors received a profile. Accuracy tells you how many of those profiles are right. A vendor can raise the first without raising the second.
What does “match rate” actually measure?
There is no standard definition. Four different numbers get called match rate:
Four things vendors call match rate
| Metric | Question it answers |
|---|---|
| Identified rate | What share of visitors received any profile? |
| Emailable rate | What share has a deliverable email address? |
| Accurate rate | What share of profiles belongs to the person who actually visited? |
| Activatable rate | What share can you use in your channels after suppression, consent, and filtering? |
Each number is smaller than the one above it. A vendor quoting the first and a vendor quoting the fourth can both be honest and still be impossible to compare.
Why aren’t vendor match rates comparable?
- Different denominators. Sessions, unique visitors, and US-only visitors produce very different percentages from the same traffic. One repeat visitor is one unique but can be five sessions.
- Bots. Bots in the denominator push the rate down. Excluding them pushes it up. Ask which the vendor does.
- Non-US traffic. Consumer identity data is built on US records. A rate quoted on US traffic only will be higher than a rate on all traffic.
- Repeat visitors. Returning visitors are easier to identify. A site with loyal repeat traffic posts higher numbers than one running mostly cold paid traffic.
- Coverage vs. correctness. Most published identification rates count profiles returned, not profiles verified as correct.
- Tag load. A visitor who leaves before the tag loads can’t be identified by any vendor. Slow pages lower every vendor’s rate.
Why do B2C sites get higher match rates than B2B sites?
The 5–20% figure that shows up in most answers comes from B2B tools measuring business traffic. Consumer traffic behaves differently.
Consumers mostly browse on personal devices, from home networks, using personal email addresses they’ve kept for years across retailers and newsletters. Those emails, stored as hashed emails (HEMs), anchor person-level website visitor identification. They tie devices, browsers, and household IP history to one person.
Business visitors often browse on work devices, on corporate networks or VPNs, with work emails that change when they change jobs. That is a thinner and less stable signal for matching a person.
Is company-level (reverse IP) identification the same thing?
No. Company-level identification looks up a visitor’s IP address and returns the organization that owns it: company name, domain, industry, and size. It doesn’t return a person. It also misses visitors on home, mobile, or remote connections, because those IP addresses belong to internet providers rather than employers. Reverse-IP rates and person-level rates measure different products.
What raises or lowers a consumer match rate?
| Factor | Effect |
|---|---|
| Traffic volume | Below about 1,000 monthly uniques, results vary more. Above about 5,000, performance is materially more stable. |
| Engagement and repeat visits | More sessions mean more chances to confirm an identity. Accuracy improves logarithmically with session frequency. |
| Non-US traffic | Lowers the rate on total traffic, since consumer data covers US records. |
| Mobile and Safari share | Browser tracking prevention shortens cookie life, making returning visitors harder to link across sessions. |
| Site type and page speed | Sites where visitors leave before the tag loads resolve fewer visitors. |
Visitor identification benchmarks at a glance
| Identification type | Typical match rate | What's returned | Main limitations |
|---|---|---|---|
| Company-level B2BReverse IP | 20–40% of total traffic | Company name, domain, industry, size | No individual. Misses home, mobile, VPN, and remote-work IPs. |
| Person-level B2B | 5–20%Commonly cited | Name, work email, title, company | Work devices and networks give weak signals. Work emails go stale when people change jobs. |
| Person-level B2CUS consumer traffic | 40–60%Untitled, US consumer traffic | Name, personal email, postal address, 30+ attributes | US records only. Varies with traffic volume, engagement, and tag load. |
Match rate measures coverage, not accuracy. Untitled reports 90–95% matching accuracy on production sites with enough traffic.
How accurate is person-level visitor identification?
No internet-scale identity dataset is 100% accurate. On live production sites with enough traffic, Untitled reports matching accuracy in the 90–95% range.
How are identities validated?
Untitled resolves identity in three phases:
- Verification. When the tag fires, it observes IP address, device and browser traits, cookie IDs, and timestamps. These are checked against known IP history and device-to-HEM relationships. A link forms only when the signals align deterministically.
- Enrichment. The verified identity is enriched through the identity graph. Records are validated by consensus across many signals rather than corrected one at a time. Records that drift or go stale lose confidence.
- Feedback. When records are activated and lead to conversions, or match a customer’s first-party data, that confirmation raises confidence for future matches.
The system also weights recency. A work email from a signup this week outranks a different one from a trade show two years ago. The full process is in How Identity Resolution Works.
Why would an identified contact say they never visited your site?
Usually for one of three reasons:
- Shared devices and households. Someone else on the same laptop or home network visited, and the identity resolved to another person in the household.
- Stale records. The match is the right person but the contact details are out of date.
- Low-signal matches. On low-traffic sites and single-session visits, matching on partial signals can produce extra, lower-confidence links.
People also forget visits, so treat one complaint as a data point rather than proof.
Sample size matters too. A system that is 93% accurate still has about 7% anomalies, so a 20-record spot check can show one or two mismatches and look worse than it is. Judge accuracy on large production datasets, not QA environments or small exports.
How does buyer intent work as a quality filter?
Visitors who return often and engage deeply give the system more signals to confirm who they are. Untitled scores visitors on visit frequency, recency, engagement depth, page patterns, time on site, and page relevance. Hot and Warm visitors make up roughly the top 25% of resolved visitors. Restricting high-cost activation to them is the fastest way to raise confidence.
How can I verify a vendor’s data before buying?
Use this checklist on evaluation calls:
- Denominator. Is the rate calculated on sessions, unique visitors, or US-only uniques? Are bots excluded?
- Which match rate. Identified, emailable, accurate, or activatable?
- Validation method. Deterministic, probabilistic, or both? Which signals must agree before a match is returned?
- Email verification and bounce rate. Are emails verified before delivery? What bounce rate do customers see?
- Refresh cadence. How often are records updated, and how are stale attributes handled?
- Sample against your own list. Ask to see identified visitors matched against your customer file. Correctly identifying people you already know is a strong accuracy check.
- Contract terms. Look for data-accuracy language, remedies for bad records, a cancellation window, and a data processing agreement that names the controller and the processor.
How do I run a fair side-by-side test of two visitor identification tools?
- Put both tags on the same site for the same month. The same window removes seasonality and promotions as variables, and a full month gives smaller sites enough sessions to stabilize.
- Use the same pages and traffic. Load both tags site-wide or on the exact same pages. Don’t give one vendor checkout and the other the blog.
- Dedupe. First remove existing customers and subscribers, since you already know them. Then split the rest into identities found by both vendors, by only A, and by only B. The unique identities are what each vendor actually adds.
- Send through the same flows. Same ESP, same email and ad flows, same timing. Send overlapping identities once so nobody gets two emails, and tag every contact by source.
- Judge by orders and revenue. Count orders and revenue from identified visitors, by vendor, over the same attribution window. Identity count is the least useful number in the test.
- Check bounce rate and complaints. Track bounces, spam complaints, and unsubscribes by source. A vendor that adds identities at the cost of sender reputation costs more than it earns.
Check out our guide on the best website visitor tracking software. Also review Untitled vs Opensend and Untitled vs Customers.ai.
Privacy and consent when running two tags
Two tags means two vendors processing your visitor data. Before the test:
- List both vendors in your privacy policy and cookie disclosures.
- Make sure your consent banner and opt-out handling cover both tags, including signals such as Global Privacy Control where state law requires them.
- Sign a data processing agreement with each vendor. In Untitled’s model, you are the data controller and Untitled is the processor.
- Remove the losing vendor’s tag when the test ends.
Is a higher match rate always better?
No. Every activation carries two risks: paying to reach the wrong person, and failing to reach a real in-market visitor. In low-cost channels like email and retargeting, where an attempt costs fractions of a cent to a few cents, missing a real buyer is usually the bigger risk, so broad activation makes sense.
In high-cost channels like direct mail, a wrong record costs real money. Filter to high-intent visitors, verify addresses, and test small cohorts before scaling.
Activation tiers by channel cost
| Tier | Channels | Measured by |
|---|---|---|
| Broad | Email, retargeting, programmatic | Aggregate ROI |
| Qualified | SDR sequences, manual outbound | Pipeline efficiency |
| Verified | Direct mail, high-cost paid social, premium sends | Strict unit economics |
Frequently asked questions
It depends on traffic type. For person-level identification on business traffic, B2B-focused sources commonly cite 5–20%. On US consumer traffic, Untitled identifies 40–60% of visitors. Compare rates only when you know the denominator (sessions, uniques, or US-only) and whether the figure counts any profile returned or only accurate, emailable ones. A good rate is one that produces orders at an acceptable cost.
Company-level identification matches a visitor's IP address to the organization that owns it and returns company details such as name, domain, and industry. It doesn't identify the individual. Person-level identification resolves the visitor to a specific person, with contact details such as name, email, and postal address. They are different products, so their match rates shouldn't be compared directly.
Consumers usually browse on personal devices from home networks and use the same personal email across many sites for years. Those hashed emails connect devices, browsers, and household IP history to one person. Business visitors often browse on work devices, corporate networks, or VPNs, with work emails that change between jobs. Consumer traffic gives an identity graph stronger, steadier signals to match.
Vendors calculate match rate differently. Some divide by sessions, others by unique visitors or US-only visitors. Some exclude bots. Some count any profile returned, others only profiles with a deliverable email. Site type, repeat-visit share, and non-US traffic also move the number. Two honest vendors can quote very different rates for the same site, so ask each for its definition and denominator.
No. Match rate measures coverage: how many visitors received a profile. Accuracy measures whether those profiles belong to the people who actually visited. A vendor can raise its match rate by accepting weaker matches, which lowers accuracy. Ask for both numbers, and test accuracy against your own customer list and against orders, not identity counts.
No internet-scale identity dataset is perfect. On live production sites with enough traffic, Untitled reports matching accuracy of 90–95%. Accuracy improves as traffic and repeat visits increase. It varies more on sites under about 1,000 monthly unique visitors, in QA environments, and in small exports. Filtering to high-intent visitors raises confidence further.
Common causes are shared devices or households, where another person on the same laptop or home network visited; stale records, where contact details are out of date; and lower-confidence matches made on partial signals, which are more common on low-traffic sites. People also forget visits. Spot checks of a few records exaggerate errors, so judge accuracy across a large production sample.
Ask for the match-rate denominator, the validation method, how emails are verified and the typical bounce rate, how often records are refreshed, and what the contract says about data accuracy. Then ask for a sample matched against your own customer list. Correctly identifying people you already know is one of the best accuracy checks available before you buy.
Yes. Install both tags on the same pages for the same period so they see the same traffic. Disclose both vendors in your privacy policy, make sure your consent and opt-out handling cover both tags, and sign a data processing agreement with each. Dedupe identities across vendors and send overlapping contacts once, so no visitor is emailed twice.
Run the test for at least one full month, with both vendors in the same calendar window. That covers weekly traffic cycles and gives identities enough sessions to stabilize. Sites under about 1,000 monthly unique visitors may need longer for reliable results. Measure orders from identified visitors over the same attribution window for each vendor.
Revenue. Identity counts reward vendors that return more profiles, including weak or duplicate ones. Orders and revenue from identified visitors, after removing existing customers and subscribers, show which vendor finds real, reachable buyers. Also compare bounce rates and spam complaints, since a vendor that hurts your sender reputation can cost more than it earns.
Bounce rate depends on whether the vendor verifies emails before delivery and how recently records were refreshed. Ask each vendor for its verification method and typical bounce rate, then measure it yourself by tagging contacts by source in your ESP. Watch spam complaints and unsubscribes alongside bounces, since all three affect sender reputation.
Person-level tools match signals from your site, such as IP address, device and browser traits, cookie IDs, and timestamps, against an identity graph. Untitled anchors identities on hashed email addresses and validates them with cookie sync partnerships, email engagement IP data, mobile device associations, and log-level timestamps. Untitled uses a consent-based framework built on US records and doesn't sell PII to third parties to make a match.
It can. Safari and some mobile browsers limit how long cookies last, which makes returning visitors harder to recognize across sessions. Visitors on slow mobile pages may also leave before the tag loads. Hashed-email and device signals help offset this. A site with a large mobile Safari share should expect some effect on its match rate.
Untitled's results vary more below about 1,000 monthly unique visitors and become materially more stable above about 5,000. Smaller sites can still use identification but should activate only Hot visitors in expensive channels. Between 1,000 and 5,000, Hot and Warm intent filtering is generally safe for one-to-one activation. Above 5,000, broader activation is more stable.
Generally yes, with proper notice and controls, but rules vary by state. Disclose each vendor in your privacy policy, honor opt-out requests and signals such as Global Privacy Control where state law requires them, and sign a data processing agreement with each vendor. This is general information, not legal advice, so confirm your setup with counsel.
About Untitled
Untitled is a consumer-first audience data platform for mid-market and enterprise consumer brands, DTC and eCommerce companies, and performance marketing agencies. It identifies anonymous website visitors as real people, builds audiences from consumer intent data, and activates them across paid social, Google, streaming TV and direct mail. It works alongside your CRM and CDP, not in place of them.
See your own match rate
The only match rate that matters is the one on your traffic. Book a demo and we’ll walk through how the Untitled ID Tag would resolve your visitors and how to set up a fair test.




