The verdict in three sentences
Most mobile-money fraud is caught with a few simple rules, no expensive AI required. In Lagos, a velocity rule (more than 3 transactions in 10 minutes) flags 60% of cases, and a SIM-swap check adds 20%. Well calibrated, this keeps a false-positive rate under 2%, so real customers are not blocked.
The rule set that catches the bulk
Fraud comes from two sources: first-party fraud (the customer themselves disputes) and social engineering (a third party manipulates the victim). A small rule set covers most scenarios without machine learning.
| Rule | What it detects | Share of fraud caught |
|---|---|---|
| Velocity > 3 txn / 10 min | Card testing, fast cash-out | ~60% |
| Recent SIM-swap check | Account takeover | ~20% |
| Device fingerprint | Multi-accounting, bots | ~10% |
| Sudden geo jump | Remotely compromised account | ~5% |
| Abnormal amount vs history | Account draining | ~5% |
The two main rules (velocity + SIM-swap) already capture 80% of fraud. The rest refine coverage.
Calibrating without smothering real customers
An over-strict rule generates false positives: legitimate customers blocked, so lost sales and an overwhelmed support desk. The goal is a rate under 2%, with a manual review queue for doubtful cases rather than an automatic block.
| Item | 2026 target | Note |
|---|---|---|
| False-positive rate | < 2% | Above that, sales are lost |
| Manual review queue | < 24h | Unblock real customers fast |
| Blocklist (numbers/devices) | Weekly update | Known repeat offenders |
| Refund-fraud pattern | Dedicated alert | Refund then dispute |
| Average chargeback cost | 5,000–25,000 FCFA | Fees + lost goods |
In Lagos, the same logic relies on a BVN (bank verification number) check and a device fingerprint: if the BVN does not match the holder, the transaction goes to review.
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Mini case study
Bimpe runs an online shop in Lagos. She faced about 50 fraud attempts per month, each of which, undetected, cost her on average 15,000 FCFA (chargeback fees plus shipped goods). By switching on the velocity rule and the SIM-swap check, she intercepts 80% of these attempts, or 40 frauds avoided. Monthly saving: about 600,000 FCFA, at a measured false-positive rate of 1.5%.
FAQ
Do you need AI to detect fraud? No. In Lagos, a few simple rules (velocity, SIM-swap, device fingerprint) catch 80 to 90% of fraud. AI only adds value at very high volume.
What is a velocity rule? It spots an abnormal number of transactions over a short window, for example more than 3 in 10 minutes. On its own it flags about 60% of fraud cases.
How do you avoid blocking real customers? By targeting a false-positive rate under 2% and sending doubtful cases to a manual review within 24h, rather than blocking automatically.
How much does undetected fraud cost? A chargeback costs on average 5,000 to 25,000 FCFA: processing fees plus shipped and lost goods. Multiplied by volume, the impact adds up fast.
What is the refund-fraud pattern? The fraudster pays, receives the product, then disputes to get a refund. A dedicated alert on fast refund requests after delivery helps spot it.
Let's talk about your project. We set up your anti-fraud rules calibrated to your volume. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
