Financial benchmarks for small US public companies, by industry

Source. The Financial Statement Data Sets published by the SEC's Division of Economic and Risk Analysis: the numbers (num.txt) and submission details (sub.txt) of every financial statement filed with the SEC in XBRL. We read the quarterly files 2025q4, 2026q1, 2026q2, 2026q3, which hold the filings made from October 2025 to September 2026.

Which companies. Each step below keeps a subset of the one before it; the count is what was left.

Why non-accelerated filers. The data set's readme lists two smaller-reporting filer-status codes, but none of the submissions in these four quarters carries either: every one is a large accelerated, accelerated or non-accelerated filer, or has no status. On today's 10-K cover page smaller-reporting status is a separate check box the data set does not record. We therefore keep non-accelerated filers (status 4-NON). Under the SEC's filer-status rules that is, broadly, a company with a public float under $75 million or a smaller reporting company with annual revenue under $100 million; a company in its first year of SEC reporting is also non-accelerated, which the revenue band bounds.

Fiscal year. A filing counts when the filer labels it fiscal 2025. 652 of the 724 companies have a fiscal year ending in December 2025; the rest end in another month. A company whose fiscal 2025 report was filed before the first quarter read is not included. Where a company filed more than one such report, the latest is used.

Industries. From the company's SIC code on EDGAR, grouped as below; the first group whose range contains the code wins, so the broad groups hold what the specific ones above them do not. Excluded: SIC 6000-6799 (finance, insurance, real estate and blank-check companies, whose statements have no comparable gross margin or working capital) and 9000 and above (shell companies and nonclassifiable establishments).

Figures read. US dollar values of standard US GAAP tags, for the whole company (no segment or co-registrant breakdown): full fiscal-year amounts for income-statement lines, fiscal-year-end balances for balance-sheet lines. Each line is read from the first tag below that the filing reports; tags are never added together. When a filing reports only one of gross profit and cost of revenue, the other is revenue minus it. Prior-year revenue, for growth, is the comparative column of the same 10-K under the same tag as this year's.

Metrics and drop rules. Each metric is computed for every company that reports its inputs, so each has its own sample. A value is dropped when an input that cannot be negative is negative (or a denominator is zero or below), or when it is above the metric's bound:

Percentiles. Linear interpolation between the closest ranks (the default of NumPy and of the spreadsheet function PERCENTILE.INC), rounded half up to one decimal place, two for ratios. Days use a 365-day year and fiscal-year-end balances, not averages.

Suppression. An industry's figure for a metric is published only when at least 30 companies reported its inputs: 83 figures are published and 27 were held back. No filing reported the same figure twice with different values.

Reproducible. The computation is a deterministic script: run again on the same four quarterly files, it writes the same CSV byte for byte.

Download the CSV (ataski-small-company-financial-benchmarks-fy2025.csv). It is the file every number on this page is read from: one row per industry and metric, with the columns industry_key, industry, metric_key, metric, unit, n, p25, median, p75. Percent metrics are in percent, day metrics in days and ratios as plain ratios.

Licence: CC BY 4.0 — free to reuse with attribution.

Two sources, two statuses. The SEC data the study is computed from is in the public domain. Ataski's dataset derived from it — the CSV and every figure on this page — is published under CC BY 4.0: you may copy, adapt and reuse it, commercially too, as long as you credit it. Attribute it as: