Identify Companies Visiting Your Website: A Working Process

To identify companies visiting your website, resolve visitor IP addresses to the organisations that operate those networks, filter out internet providers, carriers and clouds, then confirm promising accounts with behaviour or an email domain. Expect to name only a minority of B2B visits; the rest stay anonymous.

What you can and cannot see

Website visitor identification has two layers, described in BusinessMCP’s guide to how visitor identification works.

• IP to company (probabilistic). The visitor’s address maps to a network owner. It only works when traffic comes from a business network.
• Email-domain stitching (deterministic). When someone submits a form or clicks a tracked email link, the email domain confirms the company, and earlier anonymous sessions can be attached.

The guide’s figures are useful anchors: about 98 percent of visitors never fill in a form, company-level IP resolution covers roughly 20 to 35 percent of B2B traffic, and form conversion is around 2 percent. These are vendor-published ranges, not universal truths, and you should check them against your own data. We discuss why in the match rate article.

Step-by-step setup

1. Decide the question. Do you want a daily list of accounts that viewed pricing, or a monthly view of which industries read your content? The answer sets how accurate you need to be.
2. Capture the right address. Log the client IP at your edge. If a CDN or load balancer sits in front, read the forwarded header it sets, and trust it only for requests that verifiably passed through your proxy.
3. Deduplicate. Resolve unique addresses per day, not every hit. It cuts cost and noise.
4. Resolve each address. Use an API or a bulk job. BusinessMCP’s IP to company API is one option; see bulk lookups for batch handling.
5. Keep only company verdicts. Treat ISP, carrier, hosting and VPN answers as unknown.
6. Join to behaviour. Attach pages viewed, number of visits and recency to each named company.
7. Score and route. Pass accounts that fit your ICP and show intent to sales. See lead scoring with IP data.
8. Review monthly. Hand-check a sample of named accounts and adjust.

Which pages deserve attention

Not every visit is a buying signal. A sensible first pass weights these signals:

Signal	Why it matters	Caution
Pricing or docs viewed	Evaluation behaviour	Competitors and students read them too
Repeat visits across days	Sustained interest	Could be one researcher
Several pages in one session	Engagement	Bots can mimic it
Comparison pages	Active shortlisting	Often analysts, not buyers
Blog only, one visit	Low intent	Do not chase

Key takeaway: identification tells you who might be looking. Behaviour tells you whether it matters. Neither tells you who the individual is.

Filtering out the noise

The biggest source of bad data is treating infrastructure as prospects. Large carriers, broadband providers, cloud platforms and security gateways resolve constantly. A good service labels them; your job is to use the label. Filtering bot and cloud traffic goes deeper, and VPN and proxy traffic explains why many enterprise visits stay hidden.

Also exclude your own offices, your agencies and known partners, otherwise the report fills with colleagues.

Turning a name into action

A named company is a lead for research, not a mailing list. A useful routine:

• Check whether the company is already a customer or open opportunity.
• Look at the pages they read and tailor the angle.
• Send the alert to the account owner, not a general inbox.
• If a contact later fills a form, connect the two and update the record.

Warm outreach tied to real interest tends to perform better than cold lists. BusinessMCP’s guide quotes reply rates of roughly 5 to 15 percent for warm outreach against 1 to 3 percent for cold, again a vendor figure worth testing for yourself.

Privacy considerations

Company-level identification is generally easier to defend than person-level identification in the EU, but the details matter: retention of the raw address, a documented legitimate interest, disclosure in your privacy policy and consent for any tracker that stores identifiers. Our article on GDPR and IP to company lookup summarises the points, and BusinessMCP’s GDPR guide is a vendor view with its own caveats. This is not legal advice.

Tooling options at a glance

Approach	Setup effort	Output	Best for
Free web lookup	None	One address at a time	Spot checks
API on server logs	Low to medium	Named companies per day	Analysts, growth engineers
Full visitor-identification platform	Medium	Alerts, CRM sync	Sales teams
Manual registry research	High	Raw registry facts	Learning, edge cases

If you only want to test the idea, start with log exports and the free credit most pay-per-call services offer. The cheapest proof is a one-day experiment: resolve yesterday’s unique addresses, count the company verdicts and read fifty of them. Our guide on testing an IP to company API provides the method.

Mistakes that waste months

• Buying a platform before checking that your audience resolves at all.
• Counting ISP hits in match-rate claims.
• Alerting sales on every named company.
• Skipping the privacy review until after launch.
• Treating the result as a person.

Done carefully, company-level identification gives a modest but real signal about which accounts are paying attention.

A first-week plan

If you are starting from zero, this schedule keeps the effort proportionate.

Day	Task	Output
1	Export one day of unique client IPs; clean out your own traffic	A deduplicated list
2	Resolve the list with a trial account, for example BusinessMCP’s IP lookup and its free credit	Verdict counts
3	Hand-check about 50 named companies	A precision estimate
4	Join named companies to pages viewed and your CRM	A short account list
5	Review with sales; decide whether the signal is worth wiring up	Go or no-go

Only after a go decision should you automate. Many teams discover that the list is short but contains a handful of accounts they would otherwise have missed, which is enough to justify a lightweight pipeline and not enough to justify an expensive platform.

Reporting to stakeholders

Present identification as a funnel rather than a headline number:

1. Total unique visitors.
2. Visitors remaining after bot and infrastructure filtering.
3. Visitors on business networks with a named company.
4. Named companies that fit your target profile.
5. Named companies that were contacted, and how many replied.

Seeing step 3 as a fraction of step 1 sets expectations honestly, and step 5 connects the effort to pipeline, which is what the programme exists to affect. BusinessMCP’s guide makes the same point: judge the tool by pipeline generated from identified accounts, not by match-rate percentage.