What is Detection Error Tradeoff (DET)?

Detection Error Tradeoff (DET) is a curve that illustrates the trade‑off between two types of errors, false accepts and false rejects, often on normal‑deviate scales to make small differences readable. As you shift the decision threshold in a biometric or classifier system, one error decreases as the other increases. No free lunch. The curve lays that fact out explicitly.

Why it matters: product teams need thresholds that match business risk appetite. Border control picks different points than a social app. You will also track the Equal Error Rate (EER) where both errors are the same and compare curves between models or capture conditions to identify drift.

Operational practice: generate DET plots from honest test sets, segment by environment and demographics, and document the selected threshold with associated reason codes. Re‑evaluate on data drift or after feature/capture changes. In identity proofing, layer thresholding with stronger evidence for high‑risk cases—robust identity verification and liveness checks—so the system remains fair and firm even when the curve indicates trade‑offs are unavoidable.

Choose your point on the curve—own the consequences.