How Ericsson Uses Process Mining for Trade Compliance | QPR

Ericsson uses QPR ProcessAnalyzer to continuously monitor complex trade compliance flows at enterprise scale — helping its experts find the exceptional cases that matter, investigate what happened and build toward more predictive and preventive compliance.

Most conversations about process mining start with the big picture: how long things take, where work gets stuck and how close a process runs to its SLAs. At QPR Summit, Ericsson’s trade compliance team showed a very different use of the same technology. Their presentation is one of the clearest examples of what process mining can do beyond optimization.

Why trade compliance is a different process mining problem

Traditional process analysis often looks at the majority of transactions. It asks about bottlenecks, throughput and average turnaround times. For trade compliance, Ericsson often needs to do the opposite. When the team puts its “trade compliance goggles” on, it stops zooming out and starts zooming in.

In export control, sanctions and customs, the ordinary transaction is rarely the concern. The question that matters is a sharper one:

What is happening in the small number of cases that look different from everything else?

That difference might be a changed metadata field, an unusual sequence of activities, a rare combination of product, country and recipient, or a system configuration that behaves unexpectedly. Any of these can be a signal worth investigating. Often there is no bad intent at all: someone may simply do things in the wrong order through lack of knowledge, opening the door to a potential compliance problem.

Ericsson also realized early that compliance is not a linear process. It is a set of individual controls that must happen at the right points inside much larger business flows. To know whether those controls work, the team needs to understand the whole flow: where data originates, where it moves and how systems interact along the way.

From snapshots to continuous monitoring

Ericsson started its analytical journey in 2016 by redesigning its trade compliance process. Historically, much of the work had been manual, sequential and paper-based. Periodic reviews provided snapshots at particular points in time. As transaction volumes grew, that approach became increasingly difficult to scale.

The first step was mapping the business flows and the data underneath them. With that foundation in place, monitoring could be automated. Predefined criteria can flag potential issues and provide a direct link into the relevant process view, allowing the team to investigate the case without manually searching through vast transaction volumes.

The scale explains why this matters. Ericsson’s analytics environment contains more than 1.2 terabytes of transaction data, with approximately 10–20 million new transactions added every month. Monitoring that scale manually would require an impractical number of people. Machines handle the monitoring, while analysts can focus their expertise on the cases that warrant attention.

From anomaly to individual transaction

One of the most instructive parts of the presentation was a concrete example. Someone had changed the country information in a transaction document’s metadata. A predefined criterion caught the change and provided a link to a view where the difference was visible at a glance.

From there, the investigation moves from the wide view to the narrow one. First the team sees that something has changed. Then it can examine where in the operational flow the change occurred and how often the same pattern appears. When needed, analysts can drill all the way down to a single case and inspect the transaction details recorded across Ericsson’s systems.

This is where process mining earns its place. A flagged data point rarely explains itself. Seeing what happened before and after it — and how people, systems and transactions interacted around it — turns a signal into something the team can investigate and act on based on facts rather than assumptions.

Ericsson has built this into a shared analytics environment with Snowflake as the data platform, alongside case management, BI, AI and machine learning capabilities. QPR ProcessAnalyzer — QPR’s Process Intelligence platform — is the main environment for advanced process analytics; in the team’s own words, it is where they “do all the heavy lifting when it comes to analytics.” The aim is a connected analytics environment where users recognize the same numbers whether they work in QPR, Excel, Power BI or another tool.

Process mining also reveals how systems behave

One point in the presentation stood out, and Ericsson itself called it interesting: the same analytics can show how its systems are performing, not just what is happening in the process.
In a highly integrated architecture, systems depend on information propagating correctly and on time. When that does not happen, the analytics can reveal it. This gives Ericsson another way to investigate whether an issue is connected to operational behavior, system synchronization or the movement of data through the technology landscape.
For a compliance function, that distinction matters. The response to a behavioral issue may be very different from the response to an underlying system issue — and process data helps put the facts on the table.

Building toward predictive and preventive compliance

Ericsson’s ambition goes beyond explaining what already happened. The team is developing its patent-pending trade compliance analytics solution toward more predictive and preventive compliance, while also exploring how process mining, machine learning and large language models can support that journey.

The questions shift:
•    What is likely to happen, and where?
•    When is it likely to happen?
•    How much time is there to act before the point of no return?

That last question is crucial. In every flow, there is a point after which an issue can no longer be prevented and the focus shifts to controlling the impact.

As part of this journey, Ericsson is also exploring object-centric process mining to better visualize complete end-to-end business flows and understand how different elements of the process interact. At the same time, the team is investigating how machine learning and large language models can help make complex analytics easier to interpret and identify new behaviors or occurrences worth monitoring.

Getting there also requires a change in mindset: moving away from what the team calls a “desktop warrior” role — fixing things in retrospect — toward being more hands-on in the business flow and preparing for what may happen next.

Making advanced analytics easier to use

AI has another role as well: making sophisticated analytics easier for more people to use. Large process datasets can be difficult to interpret, and Ericsson is exploring how large language models and machine learning could lower that threshold, help users understand complex information and identify new behaviors worth monitoring.

A different way to think about process mining

The Ericsson story is a useful reminder that process mining is not only about making the average case faster. Sometimes its value lies in finding the exceptional case among millions of transactions — and understanding what happened around it well enough to decide what to do next.

For trade compliance, that means moving from isolated snapshots toward continuous, fact-based visibility across business flows. And as Ericsson continues its journey toward more predictive and preventive analytics, the same process foundation can support increasingly advanced ways of understanding what is happening — and what may happen next.

Continue exploring the Ericsson story

Read the full Ericsson customer story →

Watch Ericsson’s QPR Summit presentation →