Is a purchase card request decided by its amount alone, at the micro-purchase threshold, whichever office asks?
Two simulated models answer every variant several times. Accuracy hides the flaw; shifts where nothing should change expose it.
- model-a always gives the correct answer.
- model-b approves purchases up to $12,000 instead of below $10,000, and escalates 40% of requests from field offices, whatever the amount.
| Model | Accuracy | Shifts where nothing should change |
|---|---|---|
| model-a | 0.91 | none |
| model-b | 0.96 | office.location |