How the integrity score is computed

What it measures

A composite score from 0 to 100 for a procedure. 100 means no risk factor detected in the available data; a low score means signals were detected (appeals, court suspensions or annulments, severe decisions).

The four factors and their weights

• — Appeals — weight 0.25.
• — Court suspensions — weight 0.30.
• — Court reversals (annulments) — weight 0.20.
• — Decision severity (decision_content) — weight 0.25.

Sub-scores (each clamped to 0–100)

• — Appeals = 100 − (appeal count × 20)
• — Court suspensions = 100 − (suspension count × 50)
• — Court reversals = 100 − (reversal count × 50)
• — Severe decisions = 100 − (severe-decision count × 25)
• — "Severe" decisions = ProcedureCanceled, ProcedurePartiallyCanceled, RemedialMeasures, ComplaintUpheld, ComplaintPartiallyUpheld.

Aggregation and thresholds

Final score = the rounded sum of (sub-score × weight) across the four factors.

• — ≥ 80 → clean (high)
• — 50–79 → medium
• — < 50 → risk (low)
• — 100 = no risk detected.

Source and determinism

All four factors come exclusively from ANSC (appeals, decisions, court suspensions/annulments). MTender data provides context but does not enter the score. The computation is pure arithmetic: the same input always yields the same result, with no AI and no randomness.

Limitations (stated plainly)

• — If the ANSC calls partially fail, the lists stay empty and the score tends toward 100 — with an error note in the result. Always read the notes.
• — The score reflects only what exists in the ANSC data, not the intrinsic quality of the procedure.
• — The app is in a testing phase; the methodology may evolve.