Rule: the mechanism generalises, the magnitudes do not — SMBs cannot extract the same uplift Tesco / AA / Progressive did

Rule. When citing named-case magnitudes (American Airlines $1.4B, Tesco doubled share, Progressive Q1 combined ratio 86.4pct), explicitly disclose that the data volumes involved are orders of magnitude beyond a typical SMB. The mechanism generalises; the magnitudes do not.

Why. The Clubcard trial alone generated 50M transactions in three months (Tesco Clubcard 1994 trial — 9-14 stores over 3 months, >50 million transactions; dunnhumby analysed ~10pct sample); Progressive's telematics generates ~1M+ records/year per truck (Carrier Management figures referenced in the source brief); American's yield-management baseline was system-wide US passenger volume. A typical SMB lifetime transaction count fits inside any of these single-period datasets.

How to apply. For every named-case magnitude in a Candid article, include a one-sentence volume disclosure and reframe as "this is the prize if you reach this volume and these conditions." For SMB-relevant magnitude evidence, prefer alternative-data credit-scoring studies (NBER WP 29840 (Di Maggio, Ratnadiwakara, Carmichael, 2022) — "Invisible Primes: Fintech Lending with Alternative Data", J-PAL Agarwal et al. 2019 India — mobile / social-footprint ML predicts loan defaults more effectively than credit-only models) where the per-borrower mechanism is closer to per-customer SMB application.