Transaction velocity rule
A transaction velocity rule detects an account performing more transactions, or moving more total value, inside a defined time window than its configured limit allows. It looks at the rate of activity rather than any single transaction, which exposes fraud that keeps each individual amount small.
The fraud scenario this rule addresses
An attacker who has taken over an account expects a single large transfer to attract attention. Instead they send a run of smaller transfers in quick succession, each below any per-transaction amount limit, often to several beneficiaries. No individual payment is remarkable. The rate is.
The same pattern appears with stolen card details, where a fraudster makes many small purchases before the card is reported, and with mule accounts that receive and forward funds in bursts.
Data the rule requires
- Account identifier
- Transaction timestamp, from the event rather than the processing time
- Transaction amount and currency
- Transaction type or channel
- A rolling history of the account's recent transactions inside the longest window the rule uses
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 transaction_velocity
FOR EACH event E on account A:
recent = transactions WHERE account = A
AND timestamp > E.timestamp - :window
AND type IN :monitored_types
count_hit = count(recent) + 1 > :max_count
value_hit = sum(recent.amount) + E.amount > :max_value
IF count_hit OR value_hit:
RAISE alert(rule = "transaction_velocity", account = A,
evidence = recent + E)
Choosing a threshold
A velocity rule has three parameters that cannot be chosen independently: the window, the count limit and the value limit. A short window with a low count limit catches burst activity but fires on any legitimate batch. A long window with a high limit catches slow draining but reacts late, because the damage is done before the count is reached. Count and value catch different things: a count limit finds many small payments, a value limit finds a few large ones that individually pass an amount rule.
At the strict extreme the rule produces an alert for most active accounts and the queue becomes noise that analysts learn to clear without reading. At the permissive extreme it never fires, and an attacker who learns the limit can stay just below it. Because of this, institutions usually run several velocity rules at different window lengths rather than hunting for one setting that does both jobs.
Limits should come from the institution's own distribution of activity, taken per segment. A limit that suits a salary account is wrong for a merchant settlement account, so the segment and the transaction types included are part of the threshold decision, not an afterthought.
False positives to expect
Legitimate velocity spikes are common and mostly predictable: payroll runs that pay many people at once, month-end supplier payments, a customer splitting a purchase across several transfers because of a per-transaction limit, a business account batching refunds, and the first days after someone moves their main banking relationship and sets up many standing payments.
Segmenting the rule by account type, and measuring against the account's own usual rate instead of one global limit, removes most of this noise without weakening the rule for the accounts it is aimed at.
Investigating an alert from this rule
- Compare the burst with the account's history: has it ever moved this many payments or this much value in a comparable window?
- Check whether the beneficiaries are established payees or newly added, since a burst to new payees points to takeover rather than a busy month.
- Look at the device, channel and location at the start of the burst. A change that coincides with the first payment usually separates takeover from a legitimate spike.
- Check whether any credentials, contact details or limits were changed shortly before the burst began.
- Look at where the money went. If receiving accounts forward it onward almost immediately, treat the case as a mule network rather than a single compromised account.
- Record the outcome and tag the account so later activity is judged against what you found.
Related rules
Where this rule fits
The use case this rule supports is covered in velocity fraud, and the fraud rules engine page describes how conditions like these are expressed and versioned.
Common questions
What is a transaction velocity rule?
A transaction velocity rule is a fraud rule that counts transactions, or adds up their value, for an account inside a moving time window and raises an alert when the total passes a configured limit. It targets fraud that keeps each payment small, because the unusual signal is the rate of activity rather than the size of any one transaction.
The rule needs a window and at least one limit, either a count of transactions or a total value. Many institutions run several velocity rules side by side, each with a different window, so that both a sudden burst and a slower drain are covered.
Its main weakness is legitimate bursty activity such as payroll and month-end payments, which is why velocity limits are usually set per customer segment rather than for the whole book.
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