How Akoben Works: From Transaction Event to Case Resolution

Akoben receives a transaction event, evaluates it, and returns a decision, then routes anything that needs a closer look into an investigator workflow with a full audit trail. This page walks through those steps in the order they happen, from the moment a transaction event arrives to the point a case is resolved.

The eight-step transaction path through Akoben Receive transaction event, enrich transaction context, apply fraud rules, calculate risk, generate decision, create alert where necessary, investigate, then record resolution. 1. Receive transaction event

2. Enrich transaction context

3. Apply fraud rules

4. Calculate risk

5. Generate decision

6. Create alert where necessary

7. Investigate

8. Record resolution

Every step is written to the audit trail described below.

Receive transaction event

Akoben receives each transaction event from whichever system originates it: a payment switch, a core banking system, a mobile wallet platform, a card system, or a direct API call. On its own, the event carries the transaction's basic attributes — amount, currency, channel, and sender and beneficiary identifiers — and little else. Integrations covers how these systems connect.

The originating system matters because different channels surface different attributes. A card transaction arrives with merchant category and terminal data; a mobile wallet transfer arrives with agent and device identifiers instead. Akoben monitors payment, wallet, card and bank flows, and which attributes are available varies by channel.

Enrich transaction context

Before a rule can evaluate the transaction meaningfully, Akoben enriches the raw event with the context needed to judge it: the account's recent transaction history, the relationship between sender and beneficiary, and the device and channel attributes the calling system supplied. Enrichment is what turns a bare event into something a rule can reason about.

Apply fraud rules

The enriched event is evaluated against Akoben's active rule set: thresholds on transaction attributes, velocity rules counting events inside a time window, and behavioural checks. A behavioural check compares a transaction with what the account itself has done before, rather than against a limit set the same for every account. The fraud rules engine covers how these rules are written and maintained.

A transaction is generally checked against more than one rule at this stage. Some rules look only at the transaction itself; others look at how it fits into a wider pattern of recent activity on the account. Every rule that a transaction matches is recorded, along with the reason it matched, so nothing that fed the eventual decision is lost.

Calculate risk

Akoben combines the outcomes of the rules a transaction matched into a single risk score. A transaction that matches several independent rules, or one especially strong rule, scores higher than one that only brushes a single loose threshold. The score gives the calling system one value to act on and gives investigators a way to order a queue of open alerts.

Reducing several rule outcomes to one score is what makes the rest of the path practical. Without it, the calling system would need its own logic for weighing conflicting signals, and an investigator working a queue would have no consistent way to tell which of many open alerts to look at first.

Generate decision

The risk score becomes a decision. A transaction with a low score proceeds without interruption; a transaction with a high score is flagged for review, and the calling system receives that decision while the transaction is still being processed. This is what turns a monitoring signal into something with a consequence for the transaction itself, rather than only a record kept for later.

Most transactions clear this step without any human ever seeing them — the rules and the score exist precisely so that the transactions worth a person's attention are the small minority that actually need it.

Create alert where necessary

Where the decision calls for review, Akoben creates an alert. The alert carries the transaction itself, the rule or rules that matched, and the evidence those rules matched on, and it is placed into the appropriate case queue rather than a single undifferentiated inbox.

Routing matters here: an alert generated by a velocity rule on a mobile wallet account and an alert generated by a large-transfer rule on a corporate account are different problems, and placing each in a queue suited to the transaction it came from is what lets an investigation team divide the work sensibly.

Investigate

An investigator picks the alert up from the queue, reviews the evidence attached to it, and records notes as the investigation proceeds. Case management covers triage, assignment and escalation in full.

Record resolution

Whatever the investigator decides, the outcome — and every action that led to it — is written to an audit trail tied to the original transaction. That trail is what lets an institution show not only that a transaction was monitored, but that a named person reviewed it and reached a stated conclusion.

Why the path matters

Each step exists because the one before it is not enough on its own. A raw event has too little context to judge; an enriched event still needs rules to compare it against; a set of matched rules still needs to be weighed against each other before anyone can act; and a decision, once made, is only accountable if it is recorded. Transaction monitoring describes the rule families this path applies in more depth.

Common questions

How does transaction monitoring work?

Transaction monitoring works by receiving a transaction event, adding context from the account's history and the calling system, evaluating the enriched event against a set of rules, and combining the results into a risk score. Where the score is high enough, the system flags the transaction for review, creates an alert, and places it in a queue for an investigator. Every step, from receipt to resolution, is recorded so the decision can be reviewed later.

The steps run in a fixed order because each depends on the one before it: a rule cannot evaluate a transaction it has not received, and a score cannot be calculated before the rules that feed it have run.

What varies between transactions is which rules match and how high the resulting score is — most transactions pass through the whole sequence without ever producing an alert.

How is a fraud risk score calculated?

A fraud risk score is calculated by combining the outcomes of the individual rules a transaction was evaluated against into a single value. A transaction that matches several independent rules, or one rule with a strong signal, produces a higher score than one that only brushes a single loose threshold. The score gives the calling system one number to act on and gives investigators a consistent way to order a queue of open alerts by how urgent they are.

Scoring exists because a rule on its own only says whether a condition was met, not how serious the match is relative to everything else happening on the account. Combining rule outcomes into a score is what makes the difference between one weak signal and several signals arriving together visible to whoever, or whatever system, acts on it next.

The rules that feed the score are described on the fraud rules engine page.

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