Back to feed
Dev.to
Dev.to
7/24/2026
Reconcile before you round: lessons from a 14M-row open data study

Reconcile before you round: lessons from a 14M-row open data study

Short summary

A 14.4M-row study of France's SIRENE business registry reveals four data-quality disciplines: reconcile every aggregate against an independent total before publishing, count and disclose all excluded rows, flag legal-structure artifacts in rankings, and bound date fields to catch bad input. The authors caught real bugs — wrong workforce status codes, 102 records dated 2027+ — by following these rules rather than trusting queries that ran without errors.

  • •Reconcile aggregates against independent totals before publishing percentages
  • •Count and disclose every excluded data bucket rather than silently filtering
  • •Flag legal-structure artifacts (e.g., SCIs dominating 'real estate') in sector rankings

Generated with AI, which can make mistakes.

#ai-tools
Read full article at Dev.to

Is this a good recommendation for you?

Comments

Failed to load comments. Please try again.

Explore more

The original title is "The Database Was Fixed Months Ago. The Website Disagreed."
Dev.to
The original title is 10 words: "Two Pages Disagreed About the Same Law. Both Were Wrong."
Dev.to
The original headline is about ZWISERFIT nearly publishing inflated GitHub star counts (140+ vs actual 5) and how they caught it.
Dev.to
The PostgREST query that silently ORDER BY ctid: a Supabase week, distilled
Dev.to