What signs show a mismatch listing covers the whole job?
A short checklist for reseller catalogue staff to test whether a listing represents the full exception-handling job rather than only a discrepancy between two numbers.
Before you list a mismatch dataset, check that it represents the whole job and not only the discrepancy.
- **Paired records:** both sides of each mismatch are present, with their own identifiers, dates and amounts. - **Stated cause:** each case has a root-cause label such as timing difference, duplicate, keying error, rounding or missing handover. - **Evidence trail:** notes, messages or system events that a person would consult to reach the cause are included alongside the records. - **Policy context:** tolerances, approval limits and posting rules sit with the case, so the correct fix is defined rather than implied. - **Resolution record:** the reversal, adjustment or escalation that closed the item is recorded with its reason code and the role that approved it. - **Unresolvable cases:** some items correctly end in escalation or no change, so models are rewarded for stopping when they should. - **Generated status:** the listing says the data is generated, and notes where grounding and disclosure risk can be inspected by the buyer.
A listing that includes cause, evidence, policy and resolution lets customers evaluate models on the work itself, not on spotting a difference.
- Audience:
- Licensed resellers
- Content type:
- Checklist
- Buyer question:
- What it is and why it matters
Content release: r-a220eb6e099b243b
Related questions
- How does a mismatch appear in the systems an agent connects to?
- How do I tell if a process is a good fit for mismatch training?
- What does a billing mismatch caused by a transposed meter read look like?
- Why is a mismatch harder for an agent than a normal lookup task?
- What is a record mismatch in a business, and why does it matter for agent evaluation?
- What does a settlement-versus-ledger break look like as an environment task?