1. Source the fact
We prefer official product documentation, pricing pages, program terms and other primary sources. When a fact is unavailable or cannot be verified, the interface says so instead of guessing.
2. Record freshness
Important facts carry a verification date and source. A page is not made “fresh” merely by changing its publication date. Changes should correspond to actual evidence.
3. Separate fit from economics
Recommendation scoring is based on the user's requirements and verified product facts. Affiliate commission is not a recommendation input.
4. Explain tradeoffs
A good recommendation includes why a product fits and where it may be a poor fit. Comparisons should make uncertainty visible rather than bury it.
5. Human review for consequential changes
Automation can discover candidates, detect stale records and prepare updates. It must not bypass merchant approval, CAPTCHA, identity verification or other controls. High-risk commercial changes require human action.
6. Build original value
We intend to publish original analysis, benchmarks, tools and datasets where we can add evidence beyond what a merchant or feed already says. Automated aggregation without meaningful additional value is not our publishing strategy.