ATS adoption versus job counts: choose the right denominator
See why the provider with the most job posts may not have the most observed employers. Reproduce a six-company example and document the limits of an ATS sample.
All six companies and their counts are fictional. The percentages illustrate counting choices; they are not Greenhouse or Ashby market-share estimates and are not VendorCensus's live observation totals.
Name the thing you are counting
A company count, a board count, and a job-post count answer different questions. One employer can publish many jobs, operate several boards, or expose more than one provider. Before ranking ATS vendors, decide whether you are studying observed employer relationships, posting activity, or something else.
The following example counts a company once per observed provider. It separately counts listed job-post records after removing duplicate copies of the same post. It makes no claim about recruiter seats, hires, revenue, or all employers in a market.
A six-company teaching sample
Assume the five named provider relationships were checked in the same review window. Pebble Studio's provider remains unknown; its job-post count is also unknown. The company names below are invented.
| Fictional company | Observed provider | Listed job posts | Counting note |
|---|---|---|---|
| Northline Studio | Greenhouse | 12 | One company-provider relationship |
| Fir Brook Labs | Greenhouse | 2 | One company-provider relationship |
| Harbor Works | Greenhouse | 1 | One company-provider relationship |
| Cedar Labs | Ashby | 30 | One company-provider relationship |
| Orchard Apps | Ashby | 5 | One company-provider relationship |
| Pebble Studio | Unknown | Unknown | Keep separate; do not infer no ATS |
Calculate two different distributions
Among the five companies with an observed provider, Greenhouse appears at three and Ashby at two. The company-based proportions are 3 ÷ 5 = 60% and 2 ÷ 5 = 40%. The denominator excludes the unresolved company, so the label must say that.
The job-post total is 12 + 2 + 1 + 30 + 5 = 50. Greenhouse accounts for 15 ÷ 50 = 30% of these listed posts, while Ashby accounts for 35 ÷ 50 = 70%. The ordering reverses because one fictional Ashby-observed employer has many more posts. Nothing in that reversal establishes a broader adoption lead.
Using all six reviewed companies instead produces observed-provider coverage of 3 ÷ 6 = 50% for Greenhouse, 2 ÷ 6 ≈ 33.3% for Ashby, and 1 ÷ 6 ≈ 16.7% unresolved. This is a coverage breakdown with an unknown category, not a corrected estimate of the unknown company's provider.
| Measure | Greenhouse | Ashby | What the denominator means |
|---|---|---|---|
| Share of companies with an observed provider | 60% | 40% | 5 observed companies |
| Share of listed posts in this example | 30% | 70% | 50 listed post records |
| Observed coverage among all reviewed companies | 50% | 33.3% | 6 reviewed companies; 16.7% unresolved |
Resolve identities before aggregating
Two board URLs may belong to the same employer. Decide whether regional entities count independently or roll up to a parent, then apply that rule consistently. Preserve the original entity and board identifiers so a reviewer can reproduce the aggregation.
If a company exposes two providers, it can count once for each in a company-provider view. Those percentages may sum above 100%. Do not force exclusive categories unless the evidence establishes an exclusive assignment; state the overlap instead.
Post identifiers also need a definition. Greenhouse distinguishes a job-post ID from an internal job ID, while Ashby's public API distinguishes listed postings from ones intended to be reached only by a direct link. Those fields are reasons to document your inclusion rule. Counting posts is not automatically counting unique requisitions, open seats, or completed hires.
Do not turn a convenience sample into market share
A list assembled from easy-to-find public boards can overrepresent employers, regions, industries, and vendors with visible integrations. More accurate arithmetic does not repair an unrepresentative sample. Describe how companies entered the review, what was excluded, which checks failed, and the observation dates.
For a trend, revisit the same defined population with comparable collection rules. Adding ten newly researched companies can increase a provider count without a single employer adopting a new ATS. Keep newly covered entities separate from changed observations on previously covered entities.
A useful headline for the example is: Three of five employers with an identified provider showed Greenhouse, while Ashby accounted for more listed posts. An unsupported headline is: Ashby owns 70% of the ATS market. The first can be traced to the table; the second silently changes both the population and the measure.
Carry the denominator into the chart and download
Every published percentage should travel with its numerator, denominator, unit, observation window, missing-data treatment, and entity rule. Put those details beside a chart, not only in a separate methodology page. Retain the underlying rows so a reader can challenge the count.
VendorCensus's ATS list ranks providers within its reviewed sample rather than the worldwide market. Use the linked directory for its current covered records and this worksheet to plan a separate analysis. The offline worksheet stores notes only when you save an edited copy; it does not fetch boards or calculate a market estimate.