02 · Detections
Origen combines procurement data with behavioural analytics to identify hidden profit leakage, reduce procurement risk, and improve vendor performance. Every detection runs as a Postgres trigger — no scheduled jobs, no batch latency — and thresholds are tunable from Settings without a deploy.
Sample alert
"Akwasi picked Petro-Gold for diesel at GHS 22.50/unit on Tema-Aflao Phase 2 — that is 18% above Volta Petroleum's lowest bid (GHS 19.05/unit). Lost savings: GHS 4,140."
Selecting an above-market quote, especially when a cheaper compliant alternative exists, is the most common form of vendor favouritism. The variance % is configurable in Settings.
Sample alert
"Akwasi has selected Petro-Gold on 5 of the last quotes in 90 days. Pattern review recommended."
Repeat awards to the same vendor — even when each individual price looks fine — can indicate kickbacks or undisclosed relationships. The window and award-count threshold are tunable.
Sample alert
"Akwasi has raised 4 emergency POs this week (total committed GHS 89,500). Emergency procurement bypasses competitive bidding — investigate justification."
Emergency POs skip the normal RFQ + multi-quote workflow. A burst of them from one PM is a classic pattern for circumventing controls.
Sample alert
"Volta Lumber has submitted 3 invoices at ~92% of the GHS 10,000 approval threshold within 24 hours. Pattern indicates split-billing."
Splitting a large invoice into several below-threshold invoices is the most precise way to bypass approval requirements. Catching this is the difference between a clean ledger and a compromised one.
Sample alert
"Volta Petroleum submitted an invoice for GHS 22,000 matching 1 prior invoice at the same amount within 72 hours. Possible duplicate billing."
Whether it's an honest re-submission or a deliberate attempt at double payment, the cost of investigating is much smaller than the cost of paying twice.
Sample alert
Reported in the executive overview — every confirmed rebate posts to the transparent ledger with the PM's share auto-calculated, so off-book volume bonuses never disappear into nowhere.
Margin leakage is the slow drift, not a single event. Origen shows aggregate price-variance trends per vendor and per PM so leadership can act on patterns before they become a number on a write-down.
Notes for buyers
Batch jobs catch fraud after the money moves. Triggers catch it in the same transaction that records the event — so an invoice flagged as suspicious can be auto-held the moment it lands, before finance ever opens the payable queue. Median fire latency in production is under 50ms.
Each alert carries the full evidence JSON: counts, thresholds, variance percentages, vendor names, PM names, dollar amounts, and a pointer back to the row that fired the trigger. Auditors can rebuild the chain of events without ever opening a SQL console.