How Does Process Mining Work?

Process mining step-by-step

Let’s take a look at each of the basic steps involved in how Celonis process mining® works.

Modern business processes are incredibly complex. They run across a variety of systems that use different types of data, have different users, and belong to different departments.

Most process mining tools provide a few different methods for getting this data from the underlying systems. In its most basic form, a user can simply export an event log from a system as a .csv and upload it to the process mining tool. But the most advanced organizations are using real-time data ingestion to continuously sync process data.

As individual business objects, or cases, within a process move through systems, they leave a trail of breadcrumbs — whether it’s an invoice making its way from “requested” to “paid,” or a customer service ticket going from “submitted” to “resolved.” Throughout the stages of the cycle, these cases leave behind digital footprints, which are called events.

Process mining technology picks up these breadcrumbs by ingesting system logs of these events, or event logs, and visually reconstructing what’s happening in the process.

These event logs contain, at a minimum, three key pieces of process data for every individual event:

• Case ID - A unique reference to each business object
• Activity - The stage of the process the case went through
• Timestamp - The time that that specific case went through that stage

These event logs feed directly into the Process Data + Business Knowledge layer within the Context Model.

Additional information can be added as well to increase context: think vendor details for an invoice, or priority level for a service ticket.

Now that we’ve gathered all of our data, the magic of process mining can start to happen.

Process mining uses those event logs to produce an end-to-end visualization of the process that superimposes every step that every case took as it moved through the cycle into one visualization. This digital twin of the organization is a core component of the Celonis Context model that combines hindsight, insight, and foresight.

It’s a chronological sequence of events that shows all of the different paths that cases took to get from the beginning of the process to the end. Each unique path is called a variant, with variants that don’t follow a standard or accepted path being called deviations.

Process discovery allows you to explore this interactive process map and see all of the different variants.

You can zoom out and see all of the different variants, which often number in the hundreds or thousands — we affectionately refer to this as the “spaghetti” diagram.

Or, you can zoom in on the happy path, which is the way we would expect the process to look and, hopefully, the most common variant.

You can slice and dice the process anyway you’d like to see how it’s really running.

Process analytics allow you to understand the root causes of process inefficiencies, and to quantify their impact on your KPIs.

You can start to get the answers to questions about the causes of process inefficiencies:

• Where are the bottlenecks?
• What’s slowing down my throughput time?
• Which resources are overloaded?
• Which activities are often skipped?
• Which resources create deviations?

Or about the impact of process variants:

• How does a certain variant affect a process KPI like operational cost?
• How does automation, or lack thereof, impact the process cycle time?
• How does a specific business action (e.g. customer credit block) impact a business KPI (e.g. on-time delivery rate)
• What percentage of cases are conforming to the established process?

If you have assumptions about the inefficiencies, you can verify them. There are views and components that allow you to filter, slice and dice, and drill down into the data.

Or you can explore the process as-is and allow Celonis to automatically surface the most impactful inefficiencies for you using AI and machine learning.