The honest answer to “how accurate is IP-to-company identification?” is that no single number exists, and anyone who tries to identify companies from IP address accuracy alone is measuring the wrong thing. Accuracy is a property of the network the visitor used. A session that leaves through a company-owned block is a strong signal; a session from a home broadband line two streets away is not.
That matters because the second kind of session is now the normal one. Teams test a visitor identification tool on office traffic, where results look clean, then deploy it against a pipeline full of remote, mobile and VPN sessions and wonder where the matches went.
Here is what an IP address can prove, which traffic types you can name, and how confidence levels let you act on a match without overclaiming.
What an IP address can and cannot tell you
An IP address is a routing label, not a company name. What makes it useful is ownership: regional internet registries publish which organization holds each block, and routing data shows how that block behaves.
It can tell you the organization that owns the block, the network type (corporate, consumer ISP, mobile carrier, hosting or anonymizing service), and coarse geography.
It cannot tell you the person, team or department; whether the visitor sits in an office or at a kitchen table; or whether they consented to be contacted.
That gap is why privacy-first handling matters more than a bigger database. You can name a company from its own network. You cannot honestly name one from a consumer connection.
Match rates by traffic type
Quality tracks one variable: whether the visit egresses through a network the company controls.

| Traffic source | Whose network it is | Company name available |
|---|---|---|
| Corporate office or owned data center | The company | Yes, with evidence |
| Corporate VPN gateway | The company, often hosted in a cloud range | Sometimes — needs the gateway mapped |
| Coworking or shared building | The building operator | Sometimes — several tenants share it |
| Public Wi-Fi (hotel, airport, café) | The venue’s ISP | Rarely |
| Residential broadband (remote employee) | The consumer ISP | No |
| Mobile carrier (4G/5G, behind carrier-grade NAT) | The carrier | No |
| VPN, proxy, Tor, datacenter | The provider | No — and it should be flagged |
The further a visit travels from infrastructure the company owns, the less the IP address can tell you. Pretending otherwise is how teams end up emailing a domain because someone opened their pricing page from a hotel lobby.
Why remote and hybrid work lowered the ceiling
Reverse-IP coverage was built for a world where most B2B browsing happened on a corporate LAN. Remote work broke that in three ways. The egress point moved, so traffic leaves through a consumer ISP whose block belongs to the ISP, not the employer. Carrier-grade NAT puts many mobile subscribers behind one public address. And corporate VPNs increasingly terminate in cloud infrastructure, which looks like a hosting provider unless someone has mapped that gateway to the account.
Your identifiable share of traffic is therefore a subset of what it once was, concentrated on office networks and direct corporate connections. That subset is often the most valuable part of the funnel, but it is not the whole funnel.
Confidence levels turn a match into a decision
A company name without a confidence level is a guess wearing a suit. Ask for three things per match: the company, a confidence score and the evidence behind it. Then write routing rules you can defend.

- High confidence, strong ICP fit. Send to the CRM, notify the account owner, add the account to a targeted playbook.
- Medium confidence. Enrich the record, watch repeat visits and hot pages, wait for a second signal before a human reaches out.
- Low confidence or anonymous. Keep it in aggregate analytics. Do not guess a company, and never invent a person.
Reveal returns the company, a confidence score and the evidence for each match, and leaves home broadband, mobile, hosting and VPN visits anonymous instead of padding the count. Our notes on accuracy at https://pryntid.com/accuracy explain how we score it.
Where identification ends and detection begins
Anonymous does not mean uninteresting. A residential visit can be a real buyer on a high-intent page; a datacenter visit usually is not. Separating those two is a detection problem, not an identification one.
Identify covers that gap with device fingerprinting, bot and AI-agent detection, and VPN, proxy, Tor and datacenter signals, so you can decide what an anonymous session deserves: a self-serve path, a demo form, or nothing. Signal details are at https://pryntid.com/vpn-proxy-detection and https://pryntid.com/device-fingerprinting.
A policy you can ship this week
- Segment before you measure. Split traffic by network type, or you will mix office and home sessions into one misleading number.
- Set thresholds per playbook. An enterprise motion can justify a lower confidence bar than a self-serve trial.
- Keep people out of it. Link individuals only from first-party data — forms, logins, CRM imports — with a consent record.
- Publish an opt-out and keep a suppression list. It costs little and answers most privacy questions before they arrive.
- Review monthly. Compare named accounts against closed-won deals to see whether your threshold catches buyers or noise.

FAQ
Why does a visitor from a known customer show as anonymous?
Most likely they are on home broadband, a mobile connection, or a VPN that terminates in a cloud range. In each case the IP address belongs to the network provider, not the company, so a privacy-first tool has nothing defensible to name.
Can I raise match rates without guessing?
You can improve the quality of what you name — mapping customer VPN gateways, enriching from the company’s own website, confirming identity through forms and logins — but you cannot recover a company name from a consumer ISP. Treat a promise of that as a red flag.
Is IP-to-company identification a privacy issue?
IP addresses are commonly treated as personal data under laws such as the GDPR, so what matters is how you handle them. Naming a company from its own network, linking people only with consent and honoring opt-outs is defensible; guessing at individuals is not. Ask your counsel about your jurisdiction.
See what your own traffic looks like
Reveal’s free plan names up to 100 companies a month on one site with seven days of history — enough to see how your traffic splits between identifiable and anonymous accounts. Start at https://pryntid.com and check a week of matches against your CRM before you write a routing policy.
Try it free
Prynt is device intelligence with a free plan: 5,000 identifications a month, visitor IDs with confidence and bot detection — no credit card. Upgrade to Pro for VPN, proxy, tampering and behavioral signals.