Procurement Anomaly Detection

Procurement is where most public-sector corruption lives, and the patterns are statistical. Contract values that don't follow Benford's Law, a single vendor capturing an outsized share of spend, and purchase orders clustered just under the R 500 000 single-quotation threshold to dodge competitive bidding — these are the fingerprints of manipulation, and they are detectable in the data long before a whistleblower comes forward. The Treasury SCM regulations and PRECCA make these patterns matters of irregular expenditure and corruption.

Manual procurement review cannot find them. An auditor eyeballing a contracts register sees individual transactions, not the distribution; a split-PO scheme spread across months is invisible to the human reader; vendor concentration only shows up when someone totals it. So the schemes run for years, the irregular expenditure compounds, and AGSA's SCM review eventually surfaces what analytics could have caught at quarter one.

Forensic-grade anomaly detection runs the statistics over the SCM contracts register continuously — Benford's-law goodness-of-fit on contract values, vendor-concentration analysis, and split-PO detection around the threshold — and each flag carries the math behind it. Saved scans become audit evidence, and flagged patterns feed straight into the forensic case register for investigation.