Fraud Case Management with Queue Ageing and SLA Tracking

Case management is where an alert becomes a decision. Akoben's transaction monitoring evaluates each event it receives, and when rules match it raises an alert. Case management takes over from there: the alert lands in a queue, a named investigator takes it, evidence and notes accumulate, the case moves through a workflow, and an outcome is recorded. This page describes that path (the full route of an event is here), and one capability in particular. Akoben tracks how long cases have been open against a service level you configure, and shows you the ones that are overdue.

Case ageing and SLA tracking

A fraud operations manager is measured on a small set of quantities: how old the queue is, how many cases are past their target, and how much work the team clears. Akoben tracks the first two directly. Every case has an age, and each queue, or each alert type, carries a service-level target that you configure. A case that exceeds its target is a breaching case, and breaching cases are surfaced so they can be picked up first.

The target is set where the work is organised. Alerts that suggest an account is being emptied need a quicker response than a low-risk review, so they belong in a queue, or under an alert type, with a tighter target. A team handling a different product line can run to a different one. You are not forced to apply a single clock to every alert, which would make the tight targets unrealistic and the loose ones meaningless.

The value of surfacing breaches is that it turns ageing into a working list. Rather than discovering at month end that a category of alerts waited too long, a manager sees the overdue cases while they can still be reassigned, escalated or cleared. That is the difference between reporting on delay and managing it. Ageing figures also feed the productivity view in analytics and reporting, so queue performance can be reviewed over time as well as acted on today.

Alert queues

Alerts do not sit in one undifferentiated list. They land in queues, and a queue is a defined pool of work that a team or a set of investigators draws from. Queues give an operation its structure: higher-risk alerts can be routed to a queue worked by experienced investigators, routine reviews to another, and specialised patterns to the people who understand them.

Queues also give ageing a meaning. The age of a queue is the age of its oldest waiting case, and a growing queue is the earliest sign that alert volume has outrun capacity, whether because a new rule is noisy or because staffing has fallen behind. The queue is the object a manager watches and the object whose service-level target is set, which is why the structure and the tracking sit together.

Assignment to named investigators

A case is assigned to a named investigator. It has an owner, rather than being a shared item that anyone might have looked at. The owner is accountable for moving it forward, and everyone else can see who that is.

Named assignment matters for two reasons. The first is practical: two people do not duplicate the same investigation, and an unowned case is visible as unowned. The second is accountability. When a case is later questioned, the record shows who investigated it, what they recorded and what they decided. Assignment is also what makes workload measurable. The count of cases per investigator, and how quickly each is worked, is the basis of the productivity reporting described on the analytics page.

Notes and evidence

An investigation is a body of reasoning, and it has to be written down while it happens. Investigators record notes on the case: what they checked, what they found, what they ruled out and why. They also attach evidence to the case, the material that supports the conclusion, so the file is complete without anyone needing to reconstruct it from separate systems.

Notes and evidence serve three readers. The investigator who picks the case up after a colleague sees what has already been done. A supervisor reviewing the decision sees the basis for it. And a person asked later to explain why an account was treated as it was has the reasoning in one place. A case that records only its outcome tells the reader what happened but not why, which is the part a reviewer most wants to see.

Status workflow

A case moves through a status workflow. Its status says where it stands, from newly raised through under investigation to resolved, and each change is a recorded step rather than an informal update. A manager reading a queue can tell at once which cases are waiting, which are being worked and which are finished.

Status is also what ageing measures against. A case is open and ageing until it reaches a resolved state, so the workflow and the service-level target reinforce each other: the target defines how long a case should take, and the status shows whether it is on course. A defined workflow keeps cases from stalling in an ambiguous middle state where nobody is sure whether someone is dealing with them.

Escalation

Some cases need more than the assigned investigator can give. The pattern may be unusual enough to warrant a second opinion, the exposure may be large enough that a senior colleague should decide, or the case may need to go to another function. Akoben lets a case be escalated, so that it moves to the people who should handle it, with its notes, evidence and status attached.

Because the escalation happens on the case and not in a separate email thread, nothing is lost in the hand-over. The person receiving it sees the same file the first investigator built, and the record shows that the case was escalated and when. Escalation also interacts with ageing: a case that waits too long for its owner is a candidate for escalation, and the breach view is where that shows up.

Tags that feed back into monitoring

Investigation is not the end of the chain. A tag applied to an account during an investigation carries forward, so later transactions on that account are enriched with it and can be scored differently. If an investigator concludes that an account is behaving as a mule account, tagging it means the next transactions on that account are treated in the light of that finding, instead of being judged in isolation as though nothing were known.

This closes the loop between the two halves of the product. Monitoring raises the alert, the investigator reaches a conclusion, and the conclusion becomes an input to monitoring. A rule can then treat a transaction on a tagged account more seriously than the identical transaction on an untagged one. The knowledge an investigation produces is retained and applied, not left in a closed case that nobody consults again.

The audit trail

Case management keeps an audit trail of the work done on a case. It records actions taken, so the history of a case shows what was done to it and in what order. That supports the review of decisions after the fact, whether the reviewer is an internal supervisor or someone outside the operation asking how a particular account was handled.

The trail is a product of the way the work is done, not an extra step. Because queuing, assignment, notes, evidence, status changes and escalation all happen on the case, they are recorded as they happen. Nobody has to write up the investigation afterwards from memory. How the platform protects its records is covered separately on the security page.

Where case management sits in the platform

Akoben is not in the payment path. Source systems send transaction events asynchronously, either by posting them over REST or by supplying scheduled batch files. Akoben evaluates them, raises alerts, and case management is where those alerts are worked. It does not act on a live transaction. Its job is to ensure that an alert is seen, owned and resolved by a person, and that the time that takes is measured.

That fits how fraud operations actually run. The recommended outcome from scoring is passed back as a recommendation, and what the receiving system does with it is that system's decision. The investigators' work happens after the alert exists, and the quality of that work, how fast and how consistently, depends on queues, ownership and ageing. Fraud prevention produces the score and the recommended outcome; monitoring and case management are the alerting, queueing and investigation surface built on top of it.

What a fraud analyst works from

An investigator opening a case sees the alert, the rules that fired and the values they matched, the account involved and the history behind it, along with the notes and evidence already attached. The work is judgement: whether the activity is genuine fraud, an unusual but legitimate customer, or a case that needs more information before anyone can say.

Case management supports that judgement by removing the clerical work around it. The investigator is not hunting for the alert, asking who owns it, or wondering whether it is overdue. Those questions are answered by the queue, the assignment and the ageing view, which leaves the analyst's time for the decision. The outcomes feed the false-positive and rule-effectiveness reports, so the effort spent on individual cases also improves the rules that produce the next alerts.

Common questions

What does a fraud analyst do?

A fraud analyst reviews the alerts that monitoring rules raise and decides what each one means. That involves checking the flagged transaction against the account's own history, the counterparty, the channel and the device it arrived from; deciding whether the pattern is fraud, a false positive, or something that needs escalating; recording what was found; and closing the case with a stated conclusion. The judgement is the job. The system narrows down what needs judging.

This is why alert volume matters more than alert accuracy taken alone. An analyst has a finite number of considered decisions in a day, and a queue that exceeds it does not get worked more quickly — it gets worked more shallowly, which is a worse outcome than a smaller queue with a few misses in it.

What audit trail does a fraud investigation need?

It needs a record of who looked at an alert, when they looked, what evidence was in front of them, and what they concluded. That is a different question from whether the transaction was monitored, which is usually taken as given. Someone reviewing a decision months afterwards is asking whether a named person made it on a stated basis, so notes, assignment history, status changes and escalations all have to survive alongside the original transaction.

The practical test is whether a case can be reconstructed by someone who was not there. If the record shows an alert was closed but not who closed it or why, the institution has a monitoring system and no defensible decision trail, which is the half that gets asked about.

What is fraud case management?

Fraud case management is the process of taking a suspicious-activity alert and working it to a recorded outcome. Alerts are placed in queues, cases are assigned to named investigators, evidence and notes are attached, and the case moves through a status workflow, with escalation where needed. Akoben adds ageing tracking against a service level, so a manager can see which cases are overdue.

The aim of case management is to make investigation repeatable. Without it, the evidence for a decision sits in individual inboxes and spreadsheets, and nobody can say who looked at what or how long an alert waited. With it, every alert has an owner, a state and a history.

It also gives the operation a measurable shape. A queue has an age, a case has a status, and an investigator has a workload, which are the quantities a fraud operations manager is accountable for.

Read the glossary entry on case management

How are fraud alerts investigated?

A fraud alert is investigated by placing it in a queue, assigning it to an investigator, and having that investigator review the rule matches and the transaction history behind it. The investigator records notes, attaches evidence, and moves the case through its status workflow to a resolution, escalating it to a senior colleague when it needs a second opinion or a different authority.

The alert already carries the rules that fired and the values they matched on, so the investigator starts from the reason for the alert rather than reconstructing it. The work is then deciding whether the pattern is genuine fraud, an unusual but legitimate customer, or something that needs more information.

Whatever the outcome, it is recorded on the case. A tag applied during the investigation carries forward to later transactions on the same account, so the conclusion informs future monitoring.

See how alerts are raised in transaction monitoring

How do you track fraud case ageing against an SLA?

Track fraud case ageing by measuring how long each case has been open against a target set for its queue or alert type, then surfacing the cases that exceed it. Targets differ because alert types differ in urgency. Akoben lets the service level be configured per queue or per alert type and shows breaching cases, so a manager can act on overdue work.

A single service level for every alert is rarely right. A pattern that suggests an account is being drained needs a faster response than a low-risk review, and a queue staffed by one team may legitimately run to a different target than one staffed by another.

Configuring the target at the level where the work is organised keeps the measure fair. The breach view then becomes a working list: the oldest and most overdue cases first, rather than a monthly report that arrives after the delay has already happened.

See the reports that measure queue performance

What is a fraud alert?

A fraud alert is a notification raised when a transaction or a pattern of transactions matches the rules configured for suspicious activity. It records which rules fired, the values they matched on, and the account involved. An alert is not a finding of fraud. It is a prompt for an investigator to review, and it is the unit of work that enters a queue.

Most alerts turn out to be legitimate activity that happened to look unusual, which is why alerts are treated as questions to answer rather than conclusions. The proportion that turns out to be genuine is what the false-positive rate measures.

An alert that cannot be resolved quickly becomes a case, with an owner, evidence and a status. The alert is what monitoring produces; the case is what an investigation produces.

Read how alerts become cases

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