The weights below are an editable desk proposal—not a market standard.
First, require satisfactory evidence on legal scope, security, substantive quality, data access and end-to-end integration. A low price cannot offset a failed mandatory gate.
| Criterion after gates | Illustrative weight | Evidence to use |
|---|---|---|
| Total cost per accepted file | 25% | Cohort cost model, including retained work |
| Material quality and evidence traceability | 25% | Independent file review and reconstruction tests |
| Retained effort and usable-file turnaround | 15% | Measured touch time and queue distributions |
| Integration and exit | 15% | Round-trip, failure-recovery and export tests |
| Surge resilience | 10% | Stress-test results and enforceable commitments |
| Commercial clarity and change control | 10% | Negotiated scope, revisions, pricing and change provisions |
Score evidence strength as well as performance. An unverified assertion is not equivalent to a passed test. Keep unresolved mandatory evidence as not passed, rather than giving it a neutral score that could disappear inside an average.
Run at least three sensitivities: lower volume than forecast, higher retained review effort, and a larger exception share. Also show the effect of implementation costs lasting longer than planned.
What should change your buying decision?
- Prefer improved internal production if its quality and economics remain competitive after realistic tooling and supervision are included.
- Prefer estimate-writing support if drafting is the isolated constraint and retained work is controlled.
- Prefer managed production if the complete-package responsibility is contractible and demonstrated reductions in retained work justify the fee.
- Prefer targeted automation if measured task savings survive verification, integration and maintenance costs.
- Fix evidence collection first if the main cause of returns precedes production.
- Add independent QA when assurance is the missing capability—not as a substitute for correcting defective work.
Bottom line: buy a demonstrably better inspection-to-acceptance process. Do not buy an attractive first-draft price, an undefined accuracy claim or a supplier category label.