Process Mining, operational intelligence

This is how process mining enables true operational intelligence

Becoming a data driven organization is nowadays the primary goal in many performance management strategy plans. Basing decisions on insights gained from actual facts (rather than hunches, speculation, séances or other more primal means) is of course highly beneficial, so instilling this culture throughout the organization should open up vast improvement opportunities. This is also the key driver behind the current hype around self-service BI; making the tools easy to use should allow more users to reap the benefits at multiple levels of the organization, not just the top floor C-suite.

At Advectas this is also reflected in our BI maturity model, where rising anywhere above level 8 requires widespread organization use of data driven scorecards:

Process mining

Yet many organizations are struggling to find use cases for the lower levels of the organization chart, even just one or two steps below the top of the pyramid. The response from BI software vendors has been sleeker and more intuitive user interfaces, but I would argue that product complexity is not the main reason for low adoption. The real problem is a lack of relevance in the insights we can gain at anything below the strategic level of decision making. This is not due to missing features or confusing analytical environments, but a result of the kind of data the tools are designed to work with.

Since the beginning of BI, the primary focus for the solutions has been to report and analyze on results – most commonly financial outcomes. While this is great for top-level follow up, this model provides very little insight that can be used on the operational level to improve performance. Let’s take some common performance measures for outbound delivery as an example:

Process mining

From this view it is clear that our Chicago branch is underperforming, and on C-suite level this information might very well be enough – we can clearly see that we need to make an angry phone call to the windy city. But in order to actually understand the underlying operational reasons for the poor customer satisfaction and delivery precision, we can’t just focus on the final outcome since the only conclusion we can draw from this is that “something needs to be done”. In order for the local delivery operation to improve, we need a much more detailed understanding of which behaviors and actions to change in order to improve performance.

Here’s where process mining opens up a lot of new doors thanks to the increased operational granularity of its event based data model. This allows us to visualize and analyze the actual operational differences between our successful and our failing cases. To start with, we can easily generate a flowchart of the customer delivery process in Chicago where the deviations from other US regions are highlighted:

Process mining

The two most striking deviations are that Chicago gets a lot more returns and claims than the other regions, and that a much larger share of the deliveries are customer pickups. Getting more returns will of course lower profitability, so this is an outcome we should analyze further. Using the influence analysis of a process mining tool like QPR ProcessAnalyzer, we can list the root causes behind the returned orders to understand what we have done differently with those orders:

Process mining

This reveals a strong correlation between customer pick-ups and returns, so now we know which specific activity we need to improve or, if improvement is not possible, try to avoid by using other delivery methods.

With this data we have enabled the local operations team to address the specific problem, thereby providing useful information for a whole new tier of the organization. There is no way this insight could have been gained from a traditional BI solution as the event or activity data simply isn’t included, and even if it was, the tools lack the algorithms and visualizations to represent event flows and do activity based root cause analysis. Process mining dynamically provides these highly workflow specific insights for any analyzed outcome, which makes it a great way to empower lots more employees with data driven decision support – a huge step closer to achieving true operational intelligence!

If you have any questions please contact me at

Peter Selberg

Peter Selberg
Jag som skriver heter Peter och har en bakgrund som ERP-konsult och lång erfarenhet av implementationsprojekt av framför allt Infor M3. Efter att ha brottats med manuell processkartläggning och konfigurering baserad på hörsägen upptäckte jag till min stora lycka process mining. Sedan dess har jag lagt all min norrländska tjurighet på att utveckla och implementera processanalys, framförallt för kunder inom tillverkning och logistik.

Alla inlägg av Peter Selberg

Ta del av resultatet från Den svenska Business Intelligence- & Data Science-studien 2019

Ladda ner studien