Rule: three thresholds before claiming an information edge — volume to clear noise, clean data, an actual decision someone will act on (and timeliness before commoditisation)

Rule. Before a Candid article or proposal claims a client has (or can build) an information-asymmetry edge, verify four threshold conditions:

1. Volume: enough observations for the signal to clear noise (Express Analytics / INFORMS — "data quality and availability can fundamentally undermine a model's reliability").
2. Quality: clean, consistent data — "garbage in" is fatal at any scale (Clubcard data-quality lesson — multiple users on one card produced false positives in mining (the "garbage in" warning from the best-documented winner)).
3. Decision: a specific inward decision the information actually changes, with a willingness/ability to act (BDC / MIT framework — "investing in digital technologies drives revenue, but transformation management capabilities drive profits" (capability ≠ outcome)).
4. Timeliness: the edge must be acted on before the data commoditises (Tesco / Computing on dunnhumby sale — "Using CRM data from loyalty cards is now common practice rather than the unique differentiator that it once used to be", Levin (Stanford) — making private information public unambiguously improves trade; the edge erodes as data diffuses).

Why. Each threshold is independently documented. Together they explain why most named-case magnitudes (Tesco, AA, Progressive) do not scale down to typical SMB volume or maturity.

How to apply. If any of the four fails, do not anchor the article on an edge claim for that decision domain. Pivot to the capability-with-caveat framing or to a different decision domain where the thresholds do hold.