Which record elements should a dataset for this workflow include?
A checklist of the record, key, policy, label and handover elements worth requesting so that each generated case is solvable, fair and scoreable by a grader.
Use this list when you describe the data you need, so the first sample covers the relationships that make a mismatch solvable.
- Both sides of each pair, with the originating system and created, posted and changed timestamps. - Match keys in several conditions: present, missing, truncated, reused and formatted differently between systems. - Cases with one cause and cases where two causes overlap and the visible gap is their sum. - Matched pairs with no fault, so false alarms can be measured. - Irrelevant but plausible neighbouring records that act as noise during search. - Policy text for tolerances, approvals and correction methods, linked by ID to each case. - A held-out root cause and an acceptable-action list for every case. - Handover artefacts such as case notes, tickets or emails in the style your process uses.
Every item is generated, so ask to see a sample and inspect it for plausibility and disclosure risk before you build on it.
Naming these elements in your request makes the first sample cover the relationships that decide whether a case can actually be solved.
- Audience:
- Agent developers
- Content type:
- Checklist
- Buyer question:
- What can be generated
Content release: r-a220eb6e099b243b
Related questions
- What would a generated freight invoice versus proof-of-delivery task contain?
- What file layout and schema would a generated case set use?
- Can a pack include policies, tolerances and escalation rules, not just records?
- What records and relationships can MirrorFirm generate for this workflow?
- What fields and file formats can a generated mismatch pack use?
- Can a release include mismatch cases that have no single correct cause?