Change weights before scored demonstrations and apply the same changes to every option. Change the test content as well: a higher audit weight is meaningless if the supplier is still shown only a risk-register scenario.
These are proposed alternatives to the default, not sector benchmarks:
| Buying situation | Increase | Offset | Result |
|---|---|---|---|
| SOX or internal-audit-led selection | C2 +4 points; C9 +2 | C1 −3; C5 −3 | More emphasis on specialist assurance and reporting; total remains 100 |
| IT-risk and continuous-evidence programme | C5 +5; C6 +2 | C1 −4; C2 −3 | More emphasis on integration behaviour and security; total remains 100 |
| Complex incumbent replacement | C3 +3; C11 +3 | C1 −4; C2 −2 | More emphasis on historical fidelity and portability; total remains 100 |
| AI excluded from scope | C8 +2 | C12 −2 | No AI feature premium; total remains 100 |
Treat the first three rows as alternative profiles rather than automatically stacking them. A healthcare or banking label should first trigger applicability and gate discussions, not mechanically generate an arbitrary weight increase.
How do you test whether the ranking is fragile?
Recalculate under the most plausible alternative profile and the downside cost case. Also lower any disputed criterion by one score point and observe the effect.
A one-point score change in a 15%-weighted criterion changes the total by 3 points. If finalists are separated by less than that, one disputed judgement may explain the result. Report that sensitivity instead of presenting the leading score as decisive.