Data quality · 2026-05-06
Data quality notes: May 2026
A transparent note on what is working, what still needs cleaning, and how corrections are handled.
What is already checked
The build process checks for missing links, zero unit prices, household per-wash rows, known product examples and category canaries. This helps prevent obvious mistakes before data is deployed to the public site.
What still needs improvement
Some products still need better grouping. Burgers should compare with burgers, supplements should sit in vitamins and supplements, mixed vegetables should not be treated as pure broccoli, and non-food products should not appear in food diet views.
Why corrections matter
Every public row includes a report link. The long-term goal is a simple correction queue where wrong prices, categories or product matches can be reviewed and fixed without waiting for a full rebuild.
Use the live comparison
The article explains the thinking. The live table shows the current captured rows and links back to retailers for verification before buying.