Real-Time Transaction Monitoring Software for Banks and Fintechs

Transaction monitoring is the continuous evaluation of financial transactions against configured rules and behavioural expectations, in order to identify activity that may be fraudulent. This is fraud monitoring — identifying and acting on suspicious transactions — which is a related but separate function from anti-money-laundering (AML) monitoring, run under its own regulatory obligations and producing its own outputs such as suspicious activity reports. Akoben performs fraud monitoring on payment, wallet, card and bank transaction flows, and routes anything that breaches a rule into an investigator workflow.

How Akoben transaction monitoring works

Akoben receives a transaction event from a payment switch, a core banking system, a mobile wallet platform or a direct API call. Each event is enriched with the context needed to judge it: the account's recent transaction history, the counterparty relationship, and the device and channel attributes supplied by the calling system.

The enriched event is evaluated against the active rule set. Rules are expressed as configuration rather than as code. The evaluation produces a risk assessment and a decision. Where the decision is to alert, Akoben creates an alert, attaches the evidence that triggered it, and places it in the appropriate case queue.

Read the full transaction path, step by step.

Real-time transaction analysis

Akoben evaluates transactions in real time, before the transaction completes, so the calling system receives a decision while the transaction is still in progress. That is what gives an operations team the chance to act while it still matters, rather than reading about it in a report the following week.

Rules and thresholds

Rules express the conditions that make a transaction suspicious. A threshold rule fires on a single attribute crossing a configured value — an amount above a limit, a transfer to a country outside an approved list, a login from a new device. On their own, single-condition rules are blunt: raise the threshold and genuine fraud slips under it, lower it and the alert volume becomes unworkable.

A composite rule requires several conditions to hold together, which is usually what separates a genuine pattern from noise. A large transaction is ordinary on its own; a large transaction to a beneficiary added an hour ago, from a device the account has not used before, is a different case even though each individual fact — the amount, the beneficiary's age, the device — might not justify an alert by itself. Combining conditions is what lets a rule react to a pattern rather than to any single attribute crossing a line.

The fraud rules engine describes how rules are written and maintained.

Velocity monitoring

Velocity rules count events inside a moving time window: transactions per hour, cumulative value per day, distinct beneficiaries per week, failed attempts per minute. Events drop out of the count as they age past the window's edge, so the rule is always looking at the most recent slice of activity rather than everything the account has ever done.

The window length is what the rule is tuned against. A short window — attempts per minute — catches activity where the pace itself is the signal, such as automated or scripted attempts. A longer window — transactions per day or beneficiaries per week — catches an account being drained gradually, where no single transaction is large enough to trip a threshold rule but the accumulation is.

This is the specific gap velocity monitoring closes: a threshold rule judges one transaction in isolation, so activity that stays under the threshold on every individual transaction passes it every time. Velocity looks at the same transactions in aggregate, where a pattern — many small transactions in a short period, funds moving to a growing number of new beneficiaries — becomes visible even though no single transaction would have raised a threshold rule by itself.

See the transaction velocity rule in detail.

Behavioural monitoring

Behavioural rules compare the transaction with the account's own history rather than a limit set the same for every account. A limit set the same for everyone has to sit low enough to catch fraud on quiet accounts and high enough not to block legitimate activity on busy ones, which makes it wrong for most individual accounts most of the time.

Comparing a transaction with what the same account has done before avoids that trade-off. An account that regularly sends large transfers is not flagged for doing so again; an account that has never sent more than a small amount is flagged when it suddenly does, even though the amount itself might sit well under a limit set for the wider population. What counts as unusual is defined relative to the account, not to every other account on the platform — which matters most at the extremes, on accounts whose normal activity sits far from the average in either direction.

Risk scoring

Rule outcomes are combined into a risk score, which gives the calling system a single value to act on and gives investigators a way to order a queue. A binary outcome — matched a rule or did not — treats a transaction that barely crossed one loose threshold the same as one that matched several rules at once, even though the two cases call for different urgency.

A score preserves that difference. A graded response is only possible where the signal itself is graded — a single match-or-no-match outcome cannot support one, whatever the calling system does with it. A score also gives an investigator working a shared queue a consistent basis for deciding which of several open alerts to look at first, rather than working through them in the order they arrived.

Alerts and case management

An alert carries the transaction itself, the rule or rules that fired, and the evidence those rules matched on — the specific amount, counterparty, device or count that crossed the configured condition. Carrying that evidence forward means an investigator opening the alert does not have to reconstruct why it was raised from scratch.

The alert enters a queue, where an investigator picks it up, works it, and records the outcome. Case management covers triage, assignment, escalation and the audit trail in full.

False-positive management

Every monitoring rule trades detection against noise. A rule tuned too tightly generates alerts the team cannot work through, and a queue that cannot be worked is functionally the same as no monitoring. Akoben reports the alert volume and outcome distribution for each rule, so a rule that produces alerts that are consistently closed as legitimate can be identified and retuned.

Analytics and reporting covers alert volumes and how outcomes are distributed across rules.

Audit trail

Every alert, assignment, note, escalation and resolution is written to an audit trail tied to the originating transaction. A supervisor reviewing a closed case is typically not asking whether the transaction was monitored — that much is assumed — but who looked at it, when, what they saw, and why they closed it the way they did.

Financial institutions are asked to demonstrate not only that a transaction was monitored, but that a specific decision was taken by a specific person for a stated reason, and that the reasoning is still there to inspect months later. A trail that recorded only the final outcome would not answer that; recording each step as it happens does.

Integrations

Akoben ingests transactions from the systems that originate them, rather than requiring transactions to be re-entered or exported into it separately. That includes payment switches, which carry the transaction as it moves between banks or card networks; core banking systems, which hold the account and customer records a transaction is enriched against; mobile wallet platforms, which originate wallet-to-wallet and cash-out activity; and a REST API for any system not covered above.

Which of these a given institution needs depends on which channels it operates. A bank typically connects its core banking system and payment switch; a mobile money operator connects its wallet platform; a fintech operating through its own product usually integrates directly against the API.

Deployment

Akoben runs as SaaS, in a private cloud, or on-premise — deployment options chosen according to an institution's regulatory, data residency and operational requirements.

Common questions

What is transaction monitoring?

Transaction monitoring is the continuous evaluation of financial transactions against a set of rules and behavioural expectations, in order to identify activity that may be fraudulent or otherwise suspicious. A monitoring system receives each transaction, compares it with the account's history and with configured thresholds, produces a risk assessment, and raises an alert when the transaction breaches a rule. Investigators then review the alert and decide what action to take.

Monitoring differs from a one-off check at onboarding. It runs for the life of the account, because an account that was legitimate when it was opened can be taken over, sold, or recruited as a mule account later. The transaction is the observable event, so the transaction is where monitoring happens.

The rules a monitoring system applies fall into a few families: thresholds on a single transaction attribute such as amount, velocity rules that count events inside a time window, relationship rules about beneficiaries and counterparties, and behavioural rules that compare a transaction with what this account normally does.

Read the complete guide to transaction monitoring

How does real-time transaction monitoring differ from batch monitoring?

Real-time monitoring evaluates a transaction while it is still in flight, so there is still time to respond to it. Batch monitoring evaluates transactions after they have settled, usually on a schedule, and can only raise an alert for investigation after the fact. Real-time monitoring surfaces a transaction while there is still time to respond to it; batch monitoring can only find it afterwards. Most institutions run both, because some patterns only become visible across a longer window.

A real-time decision happens inside the payment path itself, so it can only use the data already available at that moment. A batch run has no such constraint and can join across a far longer history, which is why some patterns only surface there.

This is also why institutions typically run both: the real-time path catches what can be judged from immediate context, and the batch path catches what only becomes visible looking backward across many transactions.

Compare real-time and batch transaction monitoring

What is behavioural analytics in fraud detection?

Behavioural analytics compares a transaction with what the account has done before, rather than with a fixed threshold. It builds a baseline from the account's own history: typical transaction amounts, usual counterparties, normal times of day, familiar devices and locations. A transaction that departs sharply from that baseline is treated as higher risk even when it breaches no absolute limit, which catches fraud that deliberately stays under fixed thresholds.

The value of this kind of baseline is that it is specific to each account. A five thousand euro transfer is unremarkable for a corporate account and highly unusual for a salary account that has never sent more than four hundred. A single fixed threshold cannot express that difference; a baseline can.

The cost is that a baseline needs history. A newly opened account has none, which is why behavioural rules are usually combined with rules that apply specifically to new accounts.

See the unusual transaction amount rule

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