Local-first evidence check

Find the claim your data can carry.

Inspect a CSV or JSON file in this browser, see the gaps before analysis, and generate language that stays inside the evidence boundary.

  1. SourceInspect what you have
  2. BoundaryName what is missing
  3. ClaimUse permitted language
  4. ActionRequest the next step

01 · Source readiness

Check the evidence before running an analysis.

This quick audit recognizes common report fields, checks whether the observation window and location are usable, and estimates whether a privacy floor would suppress grouped results. It is a readiness screen, not a statistical result.

CSV and JSON are supported. Useful fields include latitude/longitude or address, date or timestamp, hazard type, and a segment or corridor identifier.

Other evidence already available

02 · Claim compiler

Say exactly what the evidence supports.

The readiness result sets the evidence tier; you choose the public task. NearMiss supplies bounded language, forbidden shortcuts, and the next credible action. It never turns observational data into causation or an engineering prescription.