NM-012: Vendor assessment procedure

1. Rationale

Structural controls in this archive reduce risk from the market. None of them reduce risk from the specific counterparty, which is where most of the remaining variance sits. Buyers routinely spend more effort choosing a market than choosing the vendor they will actually transact with.

2. Reading order

1. Completed order count together with the period it accumulated over, since rate matters as much as total
2. Recency of positive feedback, because a profile can coast on reputation while current performance has declined
3. Handling of negative feedback, which is the highest information field on any profile
4. Whether the listing states terms at all, including dispatch windows and what happens if a package does not arrive

3. Fields carrying little information

• Aggregate rating alone, which is shaped by who bothers to leave feedback rather than by typical experience
• Uniformly perfect and uniformly brief feedback, which reads as manufactured
• Listing presentation quality, which is unrelated to fulfilment

4. First order policy

Cap the first order with any untested counterparty. A small order costs a fraction of a large one and produces exactly the same information about whether this vendor does what they said. Losses to vendors overwhelmingly occur on first orders rather than fifth ones.

5. Refusal conditions

Decline where the counterparty asks for payment release before delivery, pushes communication off the market, or applies time pressure. None of these are conclusive individually and all of them correlate with the outcomes this procedure exists to avoid.