Data accuracy
Fundz measures how often it attaches each event to the right company and publishes the result. GET /v1/data_quality returns the latest published report as JSON: every figure with its sample size, 95% confidence interval and measurement month, the method, and the limitations. It needs no API key and no account.
The same report is rendered for people at data.fundz.net.
Endpoints
https://api.fundz.net/v1/data_qualityAuthorization header.https://api.fundz.net/v1/data_quality/:versionversion and payload_sha256.curl -s https://api.fundz.net/v1/data_quality \
| jq '{version, audit_month, overall: .entity_resolution.precision.overall}'Responses are cacheable for an hour. A version that does not exist returns 404 with {"status": "not_yet_published"}, never an empty report.
What is in the response
| Field | What it holds |
|---|---|
| version | The published version number. versions lists every version with its published_at. |
| payload_sha256 | Checksum of the stored report, so a copy can be verified. |
| audit_month | The month the sample was drawn from, for example 2026-09. |
| entity_resolution | Link precision per data lane and across lanes, completeness, duplicate rate, investor links and rater agreement. |
| automated_checks | Each automated data check with its precision on a labelled gold set, and whether it may change data without a person. |
| predictions | Calibration of predicted probabilities once their horizon has closed. Horizons still open are reported as not yet measurable. |
| methodology | How the sample is drawn, labelled and scored, in plain language. |
| limitations | What the measurement cannot see. Always present. |
| reproduce | The seed, time window and ordering rule, so the same rows can be re-drawn. |
Reading a metric
Each figure is an object with estimate, ci95 (a 95% interval; the method is named in methodology), n (decided rows), as_of, label and status. status is published only when the sample holds at least min_n decided rows (30). Otherwise it is not_yet_measured with a reason and no estimate. Treat a missing estimate as unmeasured, not as zero.
audit_month, and link data.fundz.net. The labels come from AI raters, named in methodology. This is a data-accuracy audit, not a security or compliance certification.Need help with an integration?
Tell us what you are building and we will help you get the right data into it.
Contact SupportLast updated September 2026