A REVIEW verdict is a feature, not a failure
A dairy checker with only SAFE and UNSAFE has to pretend it's sure. Adding REVIEW made it honest.
One of the best decisions in No-Whey, my dairy-ingredient checker, was making REVIEW REQUIRED a real result, not an error.
SAFE and UNSAFE seem like the only answers anyone wants. They’re clear and easy to display. The problem is that ingredient data isn’t always clear:
- Names vary, and dairy hides in derivatives most people have never heard of.
- “Non-dairy” labels can still contain casein.
- Terms like “natural flavors” can mean almost anything.
- Product databases are incomplete or disagree.
If the app only has two buckets, uncertainty doesn’t disappear. It just gets hidden inside a confident-looking verdict.
REVIEW gives those cases an honest place to go, with a plain-language explanation of why the app isn’t sure. Strong evidence resolves automatically. Weak or conflicting evidence gets explained and handed back to the person.
I keep using this pattern. It’s the same idea as the review queue in my reconciliation work at my day job. A good automated system doesn’t prove itself by answering every question. Sometimes the right output is “I don’t have enough evidence to decide this for you.”
No-Whey gives ingredient information, not medical advice.