How data providers get employee counts for private companies

Firmographic data providers get employee counts for private companies from five kinds of source: the company's own statement, a filing or registry record, an open database, a count of public professional profiles, and a model. Each kind carries a different date, and some carry none. This post shows how to tell them apart, how to judge whether a count is current, and how to test both on a sample before you buy.

Where providers get the employee count of a private company

A provider gets a private company's employee count in one of two ways: it reads a figure someone stated, or it derives one. Stated figures come from the company's own pages and announcements, from registry filings and from open databases. Derived figures come from counting public professional profiles or from a model that estimates size from other information.

The five kinds behind a firmographic employee count differ in what the number is and the date it carries.

How data companies estimate the number of employees at a private company

Where no one has stated a figure, data companies derive one. Two methods appear in providers' own documentation: counting the profiles they hold that name the company as current employer, and estimating with a model, which one provider marks as modelled rather than actual. Either result is an estimate, and a well documented record labels it as one.

A profile count moves with how many staff keep a public profile; a model's output moves with its inputs and its last run. Keep the label that says stated or estimated on every row you load.

Firmographic data providers' methodology for private companies: company websites, job postings and profiles

Firmographic data providers read a private company's employee count from the company's own website when it states one, and from filings and open databases where those carry one. Job postings and public professional profiles enter as evidence of size and growth: a posting records an intention to hire, and a profile count records who lists the employer.

Read a stated figure for its wording: "more than 200 people" is a lower bound, "about 200" a rounded point, and "200 staff across our group" spans several legal entities. Read hiring data as plans: thirty open roles say nothing about how many people work there today.

What filings and official statistics state, and when

Filings give stated counts on a fixed cycle. In the US, Regulation S-K asks a registrant to describe its human capital resources, including the number of persons it employs, to the extent that is material to an understanding of the business as a whole (17 CFR 229.101). A smaller reporting company may instead give its number of total employees and of full-time employees, again to the extent material.

The annual report on Form 10-K is due 60 days after the fiscal year end for a large accelerated filer, 75 days for an accelerated filer and 90 days for a non-accelerated filer, as the SEC's Financial Reporting Manual sets out. The count then stands as the latest annual statement until the next report.

In the UK, the notes to a company's annual accounts must disclose the average number of persons it employed in the financial year, worked out from a count for each month (Companies Act 2006, section 411). A private company has nine months from the end of its accounting reference period to file its accounts (section 442), and the Companies House register shows document images free of charge. The figure is an average over a year, and its filing deadline falls 21 months after a twelve-month year began.

Official statistics count employment at a reference date and publish totals, not companies. The US Census Bureau's County Business Patterns measure employment during the week of March 12, and publish no data that would disclose the operations of an individual employer (methodology). They size an industry or a county; they give no single company's headcount.

Employee count data freshness and timestamps by source: profiles vs SEC filings

An SEC filing gives a count the company stated, in an annual report due 60 to 90 days after its fiscal year end and filed once a year. A profile count is rebuilt on the provider's own schedule and measures who lists the employer. The filing carries the company's own date; the profile count carries the date of the provider's build.

Data freshness is the age of the date a count refers to. On that test a recent profile build is fresher than a filing, and the filing is still the company's own number.

What providers document about their employee counts

Each provider below is described by what its own documentation says about the source and the date of its employee counts: the sourcing and freshness questions among the ten questions for a company data vendor.

ZoomInfo employee count methodology for private company data

ZoomInfo's page on its data sources says it monitors public filings for information on employee headcount, revenue and other attributes. It lists the public information its automated systems read, including corporate websites, press releases, news articles, SEC filings and job postings, and names data partners and an in-house research and development team among its other sources.

Crunchbase employee count methodology for private companies

Crunchbase lists employee count among the firmographics of Crunchbase Data, and its data page names five primary sources: direct input from companies, investors and partners, user engagement signals, automated ingestion, news and government filings, and analyst validation. Ask which source supplies each count, and the date it refers to, then confirm both on a sample with the test below.

Dun & Bradstreet and People Data Labs

Dun & Bradstreet's Detailed Company Profile, as its D&B Direct 2.0 API documentation sets out, records the employee total of a single entity apart from the consolidated total of a group, and gives each a reliability text, with values such as actual, estimated and modelled. The split follows the corporate hierarchy.

People Data Labs documents its employee fields in its company schema. Its size field is a range derived from the size the company reports itself on its social media profile. Its employee count is the number of profiles in its person data that show the company as a current employer, with a start date and no end date. A monthly series counts those profiles at the end of each month, up to the last full month before its latest monthly data build.

How current employee numbers are in open company databases

An employee number in an open company database is as current as the latest statement a contributor entered, and its date is the point in time entered with it. Wikidata's employees property records a company's total employees at a given point in time, lists a point-in-time qualifier as required, and asks editors to add recent years rather than overwrite older ones.

Read the point in time, not the download date. The property's guidance generally gives the most recent count preferred rank and keeps earlier years at normal rank, so one item can hold a dated series. Where a value carries a reference, open it to see where the figure was stated.

How to tell if company headcount data is current or outdated

Company headcount data is current when the date the count refers to is recent, not when the record was recently delivered. Find that as-of date on every record, then compare its age with the source's cycle: an annual filing, a monthly build, a page seen on a given day. A count with no as-of date cannot be judged current.

Three dates can sit on one record, and only the first answers the question.

• The as-of date is the day the count was true, or the period it averages.
• The observation date is the day the provider read the statement. For an undated page it is the latest day the count can be dated to.
• The record date is the day the record was last written or delivered. It dates the record, never the count.

A test to run on a sample

Ask for sample data on fifty companies you track, with every field held on each count, and run eight checks.

1. Count the records that carry an as-of date, as distinct from an update or delivery date.
2. Label each record by kind of source, and note whether the vendor labelled it or you inferred it.
3. Compute the age of each count as today minus its as-of date, and read the oldest tenth as well as the median.
4. For each US registrant in the sample, flag a count tied to a fiscal year end more than 455 days ago, a year plus the longest 10-K deadline of 90 days, and look for the later annual report.
5. For each UK private company with a twelve-month financial year, flag a count tied to a period end more than 21 months ago, twelve months to the next period end plus nine months to file, and look for the later accounts.
6. Trace ten counts to the original statement, and check both the figure and the date.
7. Check whether each count is a point or an annual average, and never compute growth from one of each.
8. Take two deliveries some weeks apart. A record whose date moved while its count stayed the same is carrying a record date, not an as-of date.

Behind all eight sits the rule of point-in-time data: a value is usable with the date it could be known.

Worked example: dating five employee counts for Acme Robotics

This example is illustrative. Acme Robotics is an invented company, and every figure below is an assumption.

Assumptions:

• Today is 7 October 2026.
• Acme Robotics is a private company incorporated in the UK, with a financial year ending on 31 December.
• Its headcount rose by 4 people every month: 158 at the end of December 2024, 162 in January 2025 and 206 in December 2025. The monthly counts are known here only because the example is invented.
• Its accounts for 2025 were filed on 20 September 2026 and disclose an average of 184 persons employed in the year.
• Its careers page, which carries no date, says it has more than 200 people. A provider first saw the sentence on 15 September 2026.
• An open database holds 158 employees with a point in time of 31 December 2024.
• Provider A delivers a range of 201 to 500 marked as modelled, on a record last updated on 1 October 2026, with no as-of date.
• Provider B delivers 171, the count of public professional profiles naming Acme Robotics as current employer at the end of August 2026.

The working, step by step:

1. The average. The twelve monthly counts of 2025 run from 162 to 206 in steps of 4. Their sum is 12 × (162 + 206) ÷ 2 = 2,208, and 2,208 ÷ 12 = 184, the figure in the accounts. The December count of 206 sits 22 people above the average, 12 per cent of it.
2. The filing date. Nine months after 31 December 2025 is 30 September 2026. The accounts arrived on 20 September 2026, inside the period.
3. The ages. Counted to 7 October 2026, as in the table below.
4. The growth. From the open database's 158 at the end of 2024 to the December 2025 count of 206 is 48 people, or 30.4 per cent. Setting the 2025 average of 184 against the same 158 gives 26 people, or 16.5 per cent. Mixing a point with an average understates the growth by 13.9 percentage points.
5. The next comparable figure. The 2026 accounts are due by 30 September 2027. Under these assumptions the 2026 monthly counts run from 210 to 254, an average of 232, and 232 against 184 is growth of 26.1 per cent: an average against an average.

Two figures are stated by Acme Robotics itself, one is a stated figure held by an open database, and two are derived. For Acme's size today, the careers page is the freshest statement, and a lower bound. Provider A's record date is six days old, but the range itself has no as-of date, so its age is unknown. Provider B's count is the most recent dated figure, and it counts profiles, so it sits beside the stated figures rather than in their place.

How we deliver stated headcounts

Employee Headcount, part of our Announcements and scale data, delivers stated headcount over time. Each row holds one count exactly as it was stated, dated as of the day it was true, with its source and a reference to the record it was read from, one row per company, date and source. Every point in the series is a figure that was stated, so growth between two points is growth between two statements.

Each company carries one stable company ID, so headcounts join to hiring, technology and announcement data on the same key. The size bands of Market Series are bands of stated employees. The fields are defined in the data dictionary, and the method in the methodology.

Frequently asked questions

What this page says about the products it names was checked against their public documentation on 7 October 2026. Product and company names are trademarks of their owners. Fokals is not affiliated with them or endorsed by them.