← Content library

Checklist / Structured data / Advanced

Structured Extraction Validation Checklist

Validation rules to apply before extracted records reach a database or a downstream model.

For Data and backend engineers

01Schema

  • Every field has a type and an explicit required/optional decision
  • Enumerated fields list their allowed values
  • Units and currencies are fields, never assumptions
  • A source_url and retrieved_at field exist on every record

02Values

  • Numeric fields are range-checked against plausible bounds
  • Dates parse to a single canonical format
  • Missing values are null with a reason code, never invented
  • Each extracted value carries the excerpt it came from

03Review

  • A sample is reviewed per page template, not per record
  • Records failing validation are quarantined rather than dropped
  • Extraction rate per template is tracked so drift becomes visible
extractionschemavalidation