How to Find a Company's Employee Count (and Why It Varies)
By Kooperativa Engineering
A company's employee count looks like one number. It is several, defined differently and measured on different dates. A figure from an annual report, a figure on a company page and a count of profiles can all be correct and still disagree by a wide margin for the same company.
Where the number comes from
For a US public company, the annual report on Form 10-K is the primary source. Regulation S-K Item 101(c)(2)(ii) requires a description of human capital resources, including the number of persons employed. Private companies publish what they choose, usually on their own site, and the number is often a round figure or a band.
Governments define size classes for their own purposes and publish the rules, which makes them useful as a reference for what a count can mean. Data providers add a third source: counts built from professional profiles, from the company page, or from both.
Three official definitions, three ways of counting
The rules below are all official, and they count different things.
| Source | What is counted | Period |
|---|---|---|
| EU SME definition (Recommendation 2003/361/EC) | Annual work units. A person who worked full-time for the whole reference year counts as one. Part-time, seasonal and partial-year work count as fractions. Staff includes employees, owner-managers and partners with a regular activity in the business | The reference year |
| US small business size standards (13 CFR 121.106) | Everyone employed on a full-time, part-time or other basis, including workers from temporary agencies. Part-time and temporary workers count the same as full-time workers. Affiliates are added | Average of the pay periods in the preceding 24 completed months |
| SEC Form 10-K (Regulation S-K Item 101) | The number of persons employed, inside a description of human capital resources. The rule prescribes no counting method | The fiscal year end |
The SEC's Investor Advisory Committee said in a September 2023 recommendation that the human capital information available is not consistent, verified or comparable across companies, which makes the 10-K employee count difficult for investors to interpret.
The EU thresholds show why the definition matters. A company is micro under 10 staff, small under 50 and medium-sized under 250, with staff counted in annual work units. A business with 56 people, 20 of them half-time, has 46 work units, so it counts as small even though 56 people work there. The US rule would count all 56.
The same company, three numbers
To see how far the numbers drift in practice, we compared three public companies. The filing figure is from each company's latest 10-K. The next two columns come from our own data, read on 21 October 2026: the staff_count on the company record, and the number of profiles in our index that list the company as the current employer.
| Company | Latest 10-K | staff_count on the record | Profiles that list it as current employer |
|---|---|---|---|
| Microsoft | About 223,000 full-time, 30 June 2026 | 231,136 | 36,060 |
| Salesforce | 83,334 employees, 31 January 2026 | 90,908 | 17,957 |
| HubSpot | 8,882 full-time employees, 31 December 2025 | 12,262 | 4,450 |
The count on the record is higher than the filing for all three: by about 4% for Microsoft, 9% for Salesforce and 38% for HubSpot. The dates differ by months, and the 10-K figures cover different workers, full-time staff in two cases. We carry the record figure as it is published at the source and do not audit it, so treat it as a size signal and not as a headcount.
The profile counts answer another question. They tell you how many people with a profile in our index say they work there. That number depends on how many of a company's people keep a profile, so the share relative to the filing varies from about 16% for Microsoft to about 50% for HubSpot and says little about the size of a company.
Which number to use
Match the number to the decision.
| Job | Use | Watch for |
|---|---|---|
| Segmenting accounts into size bands | Any source, as a range such as 51 to 200 | Companies close to a band edge can land on either side |
| A contract or legal size test | The company's own records under the legal definition | No external count is a substitute |
| Analysis of a public company | The 10-K, with its date and definition | Full-time or all workers, and the year end |
| Tracking growth or shrinkage | The same source, read again later | Compare a number with itself, not with a number from another source |
| Judging the shape of an organization | Seniority bands as ratios | Totals depend on how many people keep a profile |
Calibrate on companies that file
If most of your accounts are private, you can still get a sense of the error. Pick ten or so public companies in your target size range, divide the record figure by the figure in the filing and look at the spread. That spread is a rough error bar for the same kind of company without a filing. It is only an indication: private companies may differ from public ones, and filing dates rarely match the date the record was fetched.
Getting the numbers from the API
The company record carries staff_count and fetched_at together, so you can see the date next to the number. Headcount by seniority returns the indexed total with a breakdown by band. Company search takes min_staff and max_staff for size bands.
# staff_count and fetched_at are in the response
curl "https://kooperativa.io/api/v1/company?username=microsoft" \
-H "Authorization: Bearer ik_live_..."
# total_indexed plus c-level, vp, director, manager and individual bands
curl "https://kooperativa.io/api/v1/company/headcount-by-seniority?company_id=1035" \
-H "Authorization: Bearer ik_live_..."
# Software companies in the US with 51 to 200 employees
curl -X POST "https://kooperativa.io/api/v1/companies/search" \
-H "Authorization: Bearer ik_live_..." \
-H "Content-Type: application/json" \
-d '{"industry": "Software Development", "country": "US", "min_staff": 51, "max_staff": 200}'Where it falls short
The indexed total counts profiles and says nothing about payroll, contractors or part-time status. The API does not return a history of the count, so growth means running the same call on a schedule and keeping the results. For a public company the filing is the better source of a total, and for a legal size test no external number will do.
Frequently asked questions
How do I find how many employees a company has?
For a US public company, read the human capital section of its latest 10-K. For other companies, check the company's own site, then a data provider's company record, and note the date of each figure. Treat private-company numbers as a range.
Why do different sources show different employee counts for the same company?
They count different things on different dates. The EU counts annual work units, the US size rules average 24 months and count part-timers fully, and a 10-K states the number of persons employed at a year end. A profile count adds a fourth measure.
How many employees make a company small, medium or large?
Under the EU definition a micro enterprise has fewer than 10 staff, a small one fewer than 50 and a medium-sized one fewer than 250, counted in annual work units. US small business size standards vary by industry, so check the standard for the company's sector.
How accurate is employee count data from a data provider?
It is a signal and not an audited figure. In our own comparison of three public companies, the count on the company record ran 4% to 38% above the latest filing. Use it for bands and direction, and use filings for totals.
Get started
Try Kooperativa
One API key. Person and company enrichment, structured search, and monitors under one flat license.