Why spot-checks matter
Automated conversion pipelines are fast, but speed doesn't guarantee correctness. A recent parser update improved table handling but accidentally broke footnote extraction. Because no one reviewed a sample of the output, the issue went unnoticed for three days. A simple spot-check protocol — review 5% of each batch randomly — would have caught it in the first hour.
How to implement spot-checks
After each batch conversion, select a random subset of files and review the Markdown output against the originals. Focus on structural elements: headings, lists, tables, links. If the sample looks correct, the batch is likely fine. If you find systematic issues, halt the batch and investigate the parser configuration before proceeding.