You turned on company-level visitor identification, and now you have a list. Three hundred rows a week, company names you recognize, company names you do not, and a sales team asking whether they should call any of them.
Here is the honest answer. That list is good at exactly one job: re-prioritizing accounts you already care about. It is bad at everything else, and it is actively dangerous when someone treats a single anonymous hit as buying intent and picks up the phone. The value is in the ranking, not in the discovery.
What company-level identification actually does, and what it reliably gets wrong
The mechanism is simple. A visitor arrives, their IP address is matched against databases of IP ranges registered to organizations, and if there is a hit, the visit gets stamped with a company name. No cookie required, no form, no login.
That mechanism has three structural failure modes, and they are not bugs that a better vendor fixes.
Remote workers on residential IPs. Someone working from home browses from a home broadband connection. The IP resolves to their internet provider, not their employer. Your best-fit buyer researching you on a Tuesday afternoon from their kitchen is invisible to this method.
Large ISP and mobile blocks. Mobile traffic resolves to a carrier. Those rows are not companies, they are address ranges. If your feed is full of telecom names, you are looking at the plumbing of the internet, not your pipeline.
Shared offices, VPNs, and cloud gateways. A visitor behind a cloud security gateway can make a large enterprise look like it is browsing from the security vendor. Coworking space, corporate VPN, same problem in reverse.
There is an inverse failure too. An IP that resolves cleanly to a large enterprise office still tells you nothing about who was actually on the page.
This is why the first thing worth doing with a visitor feed is subtraction. Filter ISP ranges, crawlers, and bots out before the list is built, not after a rep has already wasted an hour on it. A visitor list that has not been cleaned is not a signal source. It is a server log with a nicer font.
Your match rate is not the number on the vendor's homepage
Ask three sources what percentage of B2B traffic gets matched to a company and you get three different answers. Published match rates for this category come almost entirely from the vendors selling it, and they do not agree with each other.
Treat that as the finding. Nobody can tell you your match rate in advance because it depends entirely on who your buyers are and how they work. Sell to field service companies in one metro and you will match well. Sell to a distributed software company and half your real buyers will show up as residential broadband.
Measure your own. Over four weeks, count how many matched rows survive after ISP and bot filtering, then how many of those are companies you would ever sell to. That second number is your actual match rate. It is usually a fraction of the headline.
A visit is not intent. The pattern is
One pricing page hit from a five thousand person company means nothing. It could be a job applicant, a competitor, a vendor doing research, a student, or someone who clicked the wrong result. In an organization that size, a single page view is statistical dust.
Three people from the same account hitting your pricing page and a product page across nine days means something. Different sessions, multiple individuals, a mix of commercial and product pages, inside a tight window. That is a group evaluating something, which is what a B2B buying committee looks like from the outside.
So define your unit of signal and hold the line on it. Ours looks like this:
- More than one distinct visitor from the same company
- Within a rolling window of about two weeks
- Touching at least one commercial page, pricing, demo, or comparison
- On an account that fits your target profile
Anything that fails those tests is a row, not a signal. Most of your feed will fail them. That is fine. You are looking for the few accounts where the pattern holds, not trying to justify the tool by acting on everything it shows you.
The routing rule: never cold-call an anonymous visit
Write this down and give it to the sales team, because without it the feed becomes a license to cold call with extra confidence.
Never do this. Do not call a company because an anonymous visit appeared. You do not know who visited, you do not know why, and "I saw you on our website" from a stranger is a bad opening that tells the buyer you are watching them without a relationship to justify it.
Do this instead. Use the signal in three situations, all of which involve a relationship that already exists.
- Re-prioritize an open deal. An account in your pipeline shows new multi-visitor activity on pricing. That deal moves up today's call list, and the rep has a reason to ask what has changed internally.
- Revive a quiet account. A closed-lost or stalled account reappears in the feed. The buyer is in market again. Reach out about their situation, not about their browsing.
- Time a touch on a named target list. You already planned to contact this account. The signal tells you when, not whether.
In all three cases the outreach stands on its own without the visitor data. The data only changed the timing and the ordering. That is the test: if the message only makes sense because you saw them on the site, do not send it.
What to pair it with so it becomes decisions, not a feed
A standalone visitor pixel is a dashboard nobody opens by week three. It becomes useful when the visit lands on the same record as everything else you know about that company.
Three joins do the work:
- Account-level lead scoring. Score the company, not just the individual. Buying committees are plural, and a score that only moves when one contact acts will miss the pattern that matters.
- Deal stage. The same visit means completely different things depending on whether the account is untouched, in evaluation, or in procurement. Late-stage pricing visits are a negotiation signal. Pre-deal pricing visits are a qualification signal.
- Form and email history on the same company. If someone from that account opened your last three emails and a colleague just read your pricing page, you have a sequence, not a coincidence.
That is the whole argument for keeping website activity, lead scoring, the company record, and the deal in one system rather than four. When the visit shows up next to the deal, a rep makes a decision in ten seconds. When it lives in a separate tab, it becomes a weekly report that gets skimmed and ignored.
Consent, legitimate interest, and when to call legal
Plain language, and then a clear boundary.
IP addresses are not automatically outside privacy law. Whether an IP address counts as personal data, and whether IP-based tracking needs consent, varies by jurisdiction and is decided by the relevant regulators and courts, not by your vendor.
Vendors commonly frame company-level identification as running on legitimate interest with a balancing test, and person-level identification as generally requiring consent in the EU. Treat that as the industry's working position, not as settled law for your situation.
Where the line sits for you depends on your jurisdictions, your traffic, and what you do with the data. That is a legal question, not a marketing one. Put it past counsel before you switch anything on for European visitors, before you move from company-level to person-level, and before you feed identified visitors into outbound. Get the answer in writing and set your consent banner and privacy notice to match it.
The twenty minute Monday review
Three hundred rows is not a workflow. A ritual is.
Once a week, twenty minutes, one person:
- Filter hard. Target profile only. Drop ISPs, carriers, crawlers, and anyone you would never sell to. Most of the list goes.
- Sort by pattern, not recency. Multiple visitors, multiple sessions, commercial pages. Recency alone promotes noise.
- Cross-check against the CRM. Open deal, quiet account, or named target? Anything that is none of the three gets no action this week.
- Pick three. Three accounts, three owners, three specific next touches with a date.
- Close the loop next week. Did those three move? If none of them ever move, your signal definition is wrong, and you fix the definition instead of adding more rows.
Three actions a week beats three hundred rows every time. The discipline is in what you refuse to act on.
Website activity in Payani AI is built this way on purpose. ISP, crawler, and bot noise is filtered out before the list is built, so what surfaces is real engagement, and those visits feed the "Payani noticed" insights alongside lead scoring, the company record, and the deal, in one place. The point was never to show you more companies. It was to tell you which three to work on Monday.
Frequently asked questions
Is website visitor identification accurate enough to trust?
It is accurate enough to rank accounts and not accurate enough to make claims about individuals. Published match-rate estimates for company-level reverse IP range widely and contradict each other, from roughly 30 to 65 percent in one analysis, 60 to 80 percent in another, and 20 to 65 percent in a third once remote work and VPN use are considered. All of those come from sources selling the category, so measure your own rate over four weeks after filtering ISP and bot traffic.
Why does my visitor list show internet providers instead of companies?
Because the method matches IP addresses to organizations, and home broadband resolves to a residential provider while mobile traffic resolves to a carrier. Those rows are address ranges, not prospects. A usable feed filters ISP, crawler, and bot traffic out before the list is built rather than leaving your team to do it manually.
Should sales cold-call a company that visited our website?
No. You do not know who visited or why, and outreach that only makes sense because you saw them browsing is a bad opening. Use the signal to re-prioritize an open deal, revive a quiet account, or time a touch on a named target list, where the outreach would stand on its own anyway.
When does a visit actually count as intent?
When more than one person from the same company visits inside a short window, touches at least one commercial page such as pricing or demo, and the account fits your target profile. A single pricing page hit from a large company is statistical dust. Three people across nine days looks like a buying committee evaluating you.
Do we need consent to track visitors by IP address?
It depends on your jurisdictions and what you do with the data, and it is a question for your counsel. The Court of Justice of the European Union held in Breyer that a dynamic IP can be personal data where the operator has lawful means to identify the person behind it, and the European Data Protection Board's Guidelines 2/2023 confirm the consent rule reaches tracking beyond cookies. Get the answer in writing before you enable anything for European visitors or move to person-level identification.
Newsletter
Get growth intelligence in your inbox.
Sharp notes on growing with AI and what actually moves revenue. No fluff, unsubscribe anytime.



