MACHINE-LEARNING

Invoice Fraud Screening Confusion Matrix

This binary matrix separates invoices that were actually clear from those identified as fraud after review. A fraud invoice predicted clear is a false negative, while a clear invoice sent to review is a false positive. Both matter, but their business costs are different. The chart gives analysts a shared view of that trade-off before they adjust a model threshold or change review capacity. It does not establish an acceptable error rate on its own. That decision requires fraud-loss estimates, analyst capacity and governance. Use a fixed evaluation period and documented ground truth when comparing successive models.

UPDATED 2026-09-25
EXAMPLEInvoice Fraud Screening Confusion Matrix
Make this diagram your own.

Open it in the AI editor with a prompt pre-filled — keep what works, change what doesn't.

CASE ANALYSIS

Scenario

Risk analysts review a screening model before changing its alert threshold.

Key decisions

  • Identify false clears: Missed fraud can carry high risk.
  • Identify false reviews: Extra reviews consume analyst time.
  • Keep labels precise: Review is not the same as confirmed fraud.

When to reuse this

Use for threshold discussion with a documented labeling process.

FAQ

Frequently asked questions

What is a false negative here?01
An actual fraud invoice predicted as Clear.
What is a false positive?02
A clear invoice sent to Review.
Can a threshold change the matrix?03
Yes. A stricter or looser threshold changes the balance of these errors.
Open this example in the editor →

Tweak it with chat, export PNG/SVG, or fork it for your own use case.