Suspicious cash-out rule
A suspicious cash-out rule detects the conversion of electronic value into cash, through an agent, an ATM or a withdrawal, in a pattern that suggests the funds are being extracted before they can be recovered. Cash-out is the point where electronic fraud proceeds leave the system, which is why it is monitored even when everything before it looked ordinary.
The fraud scenario this rule addresses
In mobile money, a fraudster who has taken over a wallet, or who receives funds from a scam victim, needs to turn the balance into cash. The most direct route is a withdrawal at an agent. A cash-out soon after a large credit, to an agent the customer has never used, in a location unrelated to the account, is a recurring shape of this fraud.
Agents themselves can be part of the scheme, processing withdrawals for fraudsters in exchange for a share.
Data the rule requires
- Withdrawal transactions with timestamp, amount and agent or ATM identifier
- Agent location and the customer's usual locations
- The account's recent incoming credits, with source and amount
- The account's history of cash-out agents and typical cash-out amounts
- Agent identifier history, to compare cash-out behaviour across an agent's customers
Example logic
Illustrative pseudocode: it compares each event with the rule's parameters and flags the ones that match. Parameters are symbolic: every value is institution-specific and has to be derived from your own transaction distribution.
RULE suspicious_cash_out
FOR EACH withdrawal W on wallet A at agent G:
recent_credit = sum(credits to A WHERE timestamp > W.timestamp - :credit_window)
share_cashed = W.amount / max(recent_credit, 1)
unfamiliar = G NOT IN agents_used(A)
far_from_home = distance(G.location, home_area(A)) >= :max_distance
score = (share_cashed >= :cash_out_share) + unfamiliar + far_from_home
IF score >= :min_score AND W.amount >= :min_amount:
RAISE alert(rule = "suspicious_cash_out",
account = A, agent = G, evidence = W)
Choosing a threshold
A cash-out rule combines timing against credits, the share of the balance withdrawn, and how familiar the agent and location are. The share and window together decide whether the rule is looking at someone emptying a wallet right after funding it, or at normal use. A tight window and a high share is precise but misses a patient fraudster who waits a day. Loosen both and ordinary users who withdraw most of their balance on payday are caught.
Distance depends heavily on the market. Where customers routinely use agents near work rather than near home, a distance condition fires on routine behaviour, and in markets with few agents, travelling a long way to the nearest one is normal.
A minimum amount stops small everyday withdrawals alerting, but a high one allows structured withdrawals just beneath it, so cash-out is usually watched at the agent level as well as the customer level.
False positives to expect
Most cash-outs are entirely legitimate. A worker paid by wallet who withdraws their wage in full, a customer paying a market trader in cash, a traveller using an agent far from home, a family receiving a remittance and withdrawing it the same day, and someone emptying a wallet before it expires from inactivity all look like the pattern.
Payday and remittance peaks create predictable spikes. Recognising salary and remittance credits as expected funding, and learning each customer's usual agents and times, are the main ways to reduce them.
Investigating an alert from this rule
- Look at the credits that funded the withdrawal: their source, amount and how long before the cash-out they arrived.
- Check whether the wallet's owner has used this agent or location before.
- Review how the session or instruction was made: the device, whether credentials or the registered phone number changed recently.
- Check whether the funding came from a reported victim, or a sender linked to other alerts.
- Review the agent: do many wallets with similar patterns cash out through the same agent?
- Contact the wallet owner, and where an agent looks implicated, escalate through the channel set up for agent oversight.
Related rules
Where this rule fits
The use case this rule supports is covered in fraud monitoring for mobile money, and the fraud rules engine page describes how conditions like these are expressed and versioned.
Common questions
How do you detect suspicious cash-out activity?
Suspicious cash-out activity is detected by comparing a withdrawal with the credits that funded it and with the account's usual behaviour. Useful signals are a large share of a recent credit withdrawn soon after it arrived, an agent or location the customer has not used before, and a distance from the customer's usual area. A rule combines several of these rather than relying on one.
Cash-out is the point where electronic fraud proceeds leave the system, so it is worth monitoring even when earlier steps looked ordinary. The rule needs the credit history alongside the withdrawal, since the withdrawal alone says very little.
Wage withdrawals, remittances and traders paid in cash all resemble the pattern, so recognising expected funding and each customer's usual agents matters. Monitoring at the agent level as well catches structured withdrawals that stay under a customer-level amount.
Discuss your transaction monitoring requirements
Tell us which transaction flows you need to monitor, which systems Akoben would integrate with, and how your fraud team works today.
Request a demo