A coupon experiment assigns offers by ZIP code, but many residents shop across ZIP-code boundaries and share coupon codes with friends. Sales rise in both assigned and unassigned areas, with a slightly larger rise in assigned areas. Which missing bias is most likely?