How to Find Information on a Company: Manual Research vs. a Company Data API

Eight pieces of information describe a B2B company well enough to qualify it: what it does, how many people work there, where it is, which industry it belongs to, who runs it, how it is funded, what technology it runs on, and whether it is hiring. All eight exist publicly. For one company, you can find them by hand in about fifteen minutes. For fifty, the method breaks — not because the information disappears, but because the fifty answers stop agreeing with each other. This article covers both paths: how to research a company manually, where each fact lives, and what changes when you need the same eight fields for every company on a list.

The eight fields, and where each one lives

• Description and offering — the company website, its LinkedIn "About" section, and its listing on product directories. The website says what the company wants to be; the directories say what customers think it is.
• Headcount — declared on the LinkedIn company page as a range; observable by counting public employee profiles; occasionally filed with a registry. These three rarely match, and the gap is informative — more on that below.
• Location — headquarters on the website and registry; offices and remote spread from employee profiles.
• Industry — self-declared on the company page, assigned by registries under national codes (NAICS, NACE, SIC), and assigned again, differently, by every data provider. Pick one taxonomy and map the others to it.
• Leadership — the team page, LinkedIn searches on title and company, registry filings for directors.
• Funding and financials — press releases and funding databases for private companies; filings for public ones; registries for the basics almost everywhere.
• Technology stack — job postings (the most honest source), the website's own code, integration marketplaces.
• Hiring activity — the careers page and job boards. Open roles are the best leading indicator of where a company is investing.

Researching one company by hand

Start with the domain. Everything else hangs off it. The website gives you the offering, the headquarters, the team, and usually a careers page. The LinkedIn company page gives you the declared headcount range, the industry label, and a link to every employee who has made their profile public. A search on the national business registry gives you the legal entity, incorporation date, directors, and — depending on the country — filed accounts. Ten minutes of press search fills in funding and recent news. Job boards tell you what the company is building next.

For a single, well-established company with a normal web footprint, this works and it is free. Budget fifteen to twenty minutes and expect to end with a coherent picture.

The headcount problem: declared versus observed

Headcount is the field where manual research most reliably misleads, and it is worth understanding why before you build anything on it.

There are two numbers. The declared headcount is what the company states — a range on its LinkedIn page, chosen by whoever set the page up, and updated when someone remembers. The observed headcount is what you can count: the number of people whose public profiles say they currently work there. Both are wrong in different ways. The declared range lags reality; the observed count misses people without public profiles and includes people who left without updating.

The gap between them moves. In a 2025 Reverse Contact study, 16.6% of companies had a headcount that moved out of its declared range within twelve months. One company in six, in a year, is no longer the size its own page says it is. For fast-growing companies the gap is the single most useful signal you can extract — a startup declaring "11–50" with 140 observable employees is telling you something its About page does not. One caveat, which applies to everything in this section: AI is changing how much output a team of a given size produces, and headcount may become a weaker proxy for company scale than it has been. The ratio between the two numbers will remain informative; what it predicts may shift.

The practical rule: never treat a declared range as a fact. Treat it as a claim, and check it against the observed count when the decision matters.

Where the manual method breaks

Around fifty companies, three things happen at once.

• Time. Fifteen minutes each is twelve hours. For five hundred companies it is three weeks of one person's work, before anyone has done anything with the result.
• Consistency. Two researchers describe the same company differently. One writes "SaaS", the other "Software Development". One records the declared headcount range, the other counts profiles. The spreadsheet now contains two incompatible definitions of every field, and a filter on either one silently drops half the list.
• Freshness. The first company you researched is three weeks older than the last. One in six has, on the headcount evidence alone, already changed category. Nothing in the spreadsheet says which.

None of these is a research failure. They are properties of doing structured work with unstructured tools.

The programmatic path: domain in, profile out

A company data API replaces the fifteen-minute routine with one request. You send a domain — or a company name, or a LinkedIn company URL — and receive a structured profile with the eight fields above and their sub-fields, in the same schema for every company, with a timestamp saying when the record was last refreshed. Reverse Contact exposes this through the Company Enrichment API; the full field list is on the Company Data page.

The response for a fictional northwind-analytics.com reads, in outline: legal name and description; industry under one normalised taxonomy; headquarters in Lisbon with two further office locations; declared headcount range "51–200" and an observed count of 184; founding year; a funding stage; a technology stack of a dozen named tools; nine open roles, six of them in engineering; the LinkedIn company ID that lets you join this record to every employee profile; and a last-updated timestamp from four days ago.

When the question runs the other way — not "describe this company" but "find every company that looks like this" — the Company Search API takes filters (industry, headcount band, location, technology, hiring) and returns the matching list, each entry resolvable to a full profile.

Freshness, confidence, and the honest null

Three design choices make the programmatic result more trustworthy than the spreadsheet, not just faster.

• Every record is dated. The last-updated field tells you whether you are looking at last week's company or last year's. A list where every row carries its own freshness is a list you can reason about.
• Every match carries a confidence score. Resolving a name to the right company is a matching problem — "Apex" is several hundred companies. The API tells you how sure it is, and you can raise the precision threshold when a wrong match costs more than a missing one.
• Unknown is null, not a guess. A field the sources do not support comes back empty. Fields we cannot verify come back null — never guessed. An empty funding field means "not publicly known"; it does not mean "bootstrapped".

What this does not cover

Three kinds of company are under-described by any method built on public professional data. Companies with almost no web footprint — a regional manufacturer with a two-page website and no LinkedIn presence — return thin profiles, honestly. Subsidiaries and brands are frequently conflated with their parents, because the public sources conflate them; check the legal entity when it matters. And domain homonyms — two unrelated companies on .com and .co — are a matching hazard that the confidence score flags but cannot always resolve. In all three cases, the manual method has the same problem; it just hides it better.

Frequently asked questions

How do I find out how many employees a company has?

Two ways, giving two numbers. The declared range is on the company's LinkedIn page. The observed count is the number of public profiles listing the company as current employer — a company data API returns both. In a 2025 Reverse Contact study, 16.6% of companies had moved out of their declared range within a year, so treat the declared figure as a claim to be checked, not a fact.

What is the best free way to research a company?

The company website, its LinkedIn page, the national business registry and the careers page, in that order. Fifteen to twenty minutes per company, reliable for one, unworkable past fifty.

What is a company data API?

A service that returns a structured, dated profile of a company — description, industry, headcount, location, leadership, funding, technology and hiring — from an identifier such as a domain, in the same schema for every company. It replaces manual research when the number of companies makes consistency and freshness the problem rather than access.

How fresh is company data?

It depends on the provider and the field; headcount and hiring change fastest. Reverse Contact returns a last-updated timestamp on every record and lets you request a live refresh when the decision needs it.