Who visited my website? Company identification without tracking people

Who visited my website? To find out, use B2B website visitor identification: it matches each visit to the company whose network it came from, and names the company, never the person. It will not name every visitor, and it should not try to.

Your analytics tool tells you that 412 sessions came from organic search last week and that the pricing page was read 38 times. It does not tell you that six of those readers worked at companies you would love to sell to, or that one of them came back three times. For a B2B company that sells to a few hundred accounts, the count is the wrong unit. The company is the unit.

"Who visited my website?" is the question every founder asks the first time they see a spike. This post explains how company identification answers it, why it can do so without tracking any individual, and what a small team should do with the list once it has one.

How website visitor identification turns a visit into a company name

When a browser requests a page, the request carries an IP address. Many companies route their office and VPN traffic through addresses that are registered to them, and those registrations are public. Identification services keep a database that maps those ranges to company names and enrich each match with public firmographics: industry, size, revenue band, the technologies the company's site uses.

That is the whole mechanism. No cookie is needed to name the company, no login, no form. The match is at the company level because the address belongs to the company, not to anyone in it. Home broadband and mobile networks resolve to the carrier, not to an employer, so a large share of visits stay anonymous. Any tool claiming to name every visitor is wrong about how the internet works.

Companies only, never people

The line that matters is between the company and the person. Knowing that someone at a 200-person logistics firm read your pricing page twice is a sales signal. Knowing which employee did so is surveillance, and in most jurisdictions it is also unlawful without consent. A well-designed identification setup stores the company, the pages, the visit count and the dates, and stores nothing that could pick out an individual.

This is how Mati handles it. One snippet on your site, and the leads page shows companies: the name, the pages they read, how many times they came, when, and how they got there. It never shows a person's name from a visit. If you want people, Mati separately offers a directory of the company's public roles you can search by seniority and department, which is a different and consented source, not the visit log.

The list is useless without fit

The first week of identification is exciting and the second is noisy. Universities, agencies checking your site for a pitch, your own competitors, a recruiter. A raw list of visiting companies wastes a small team's time. What makes it useful is a fit judgement against your ideal customer profile: the industries you sell to, the company sizes, the roles you need in the building, the disqualifiers.

Mati scores every identified company against your ICP and says which axes matched, so "fits" is a checkable claim rather than a label. A company that fits and read the pricing page is a different animal from one that fits and read the careers page. The pages tell you the intent; the fit tells you whether the intent is worth your hour.

Where they came from

Half the value of the list is attribution. A company that arrived from an article you published last month is proof that the article works. One that arrived from a LinkedIn post tells you the post reached the right audience even if it got twelve reactions. One that arrived from an AI assistant's link is the first sign your AI visibility work has started to pay.

Mati traces each visit back to the piece that brought it with three levels of confidence, and says which one applies: the company told you on a form, they arrived on a link Mati tagged in a post or article, or the source is inferred from the referring site. It reports the level rather than pretending every attribution is certain, because a monthly report that overclaims is worse than no report.

What to do with the list on Monday

• Read the fits that came back. A second visit from a company that fits is the warmest signal you own. Someone there is evaluating you.
• Look at what they read. Pricing, a comparison page, a case study: each suggests the next message. A company reading your integrations page needs an integrations answer, not a brochure.
• Find the right person, properly. Use the company's public roles to pick the seniority and department that buys, and reach them through the channels they accept. Never write "I saw you on our site"; you did not see a person, and it sounds like you did.
• Feed the plan. If three companies that fit arrived from one article, write its sibling. If none of the fits came from LinkedIn, the posts are reaching the wrong people, however well they perform.

In Mati, the list is one step in marketing run end to end: the same plan that drives your SEO articles, AI-answer work and LinkedIn posts reads it. Ask what to do next and Mati will cite the companies that came, which of them fit, and what brought them, and it will propose the work: a sequence to the people who fit at a company that returned, rehearsed before anything goes out, or the article that would bring more like it. It proposes; you apply. Nothing is sent on your behalf.