Protocol status: This is a proposed audit protocol. It reports no customer, processor, or staffing result.
Define the states being compared
Authorization, refund approval, processor initiation, settlement, and bank posting are different states. Define which payment event a customer message is expected to match at its send time.
Preserve the exact message version and processor evidence. Do not score a later successful outcome as proof that an earlier message was accurate.
Select records before scoring
Freeze the review window and eligible refund communications before reviewers see agreement results. Retain partial refunds, reversals, missing processor events, and corrected messages as declared categories.
Use study keys and keep direct customer identifiers out of the analysis table. Restrict raw message and payment access to reviewers who need it.
Create a versioned codebook
Define exact agreement for amount, currency, state, and event time. If a compatible-state mapping is useful, predeclare it as a sensitivity analysis rather than changing the primary rule after seeing results.
List allowed missing, not applicable, conflicting, and unresolved values. Unknown evidence must not be converted into agreement.
Review independently
Two reviewers should code an overlap sample without seeing each other’s decisions. Publish raw agreement, disputes, adjudication reasons, and codebook updates.
Reconstruction matters as much as the final code. A reviewer should reach the source message and payment event from stable references rather than screenshots with unclear context.
Report with limits
Show flow counts, source coverage, agreement counts, disagreement types, uncertainty, and counterexamples. Keep the original specification visible beside sensitivity results.
The protocol can reveal message-event mismatch and evidence gaps. It does not determine legal rights, individual remedies, processor quality, or the cause of an observed discrepancy.
Sources and boundaries
These sources provide public context for the protocol. Their inclusion does not mean they endorse this site or that the proposed study has been conducted.