The standard
We build our products to the standard of a methods paper. These are the rules, in the order we apply them.
1. Cite every value
Every number a product shows traces to a source: the URL, the time we retrieved it and a checksum of the file. Anyone can reproduce the pipeline from its code. When a claim rests on research, we rely on peer-reviewed work and read it in full, not just the abstract.
2. Invent nothing
We never estimate a value from memory or fill one in to close a gap. A missing value reads "Data not currently available", and we report how many are missing. If a ranking would depend on guessed data, we do not show it.
3. State the limits
Every method involves choices the evidence does not settle: a window, a threshold, a weight. We label those as design choices instead of presenting a convention as a finding. Scout goes further: every assessment lists what it could not see, and its written verdict can never claim more than the measured signals support.
4. Go to primary sources
We choose government and academic data over third-party indexes, an agency's own documentation over a summary of it, and a city's financial report over an encyclopedia entry. We keep a secondary source only when no other exists, and we mark it.
5. Check against common sense and outside benchmarks
We compare results with known facts and independent measures before anyone sees them. When a check finds an error, we correct it in the open and record the correction with its date.