A Simple Way to Detect Bottlenecks in Complex Processes

We all work with complex business processes every day, and analysing them is often difficult. As a solution Architect, I’ve been involved in discussions about many business processes that keeps breaking or that end users are unhappy with. It’s either too time-consuming or too complicated, and at first glance it feels overwhelming.

In this blog post, I’m sharing an idea for identifying bottlenecks in a process. I recommend using a concept called process mining.

I first came across process mining 6–7 years ago during an interview, and it has helped me a lot since then. Process mining is a tool that can identify bottlenecks very effectively using data. Data flows everywhere in an organisation—through Kafka events or across multiple systems connected by a correlation key. To get insights, we need three essential parameters: case ID, activity, and timestamp. Case ID (or correlation key, request ID, process ID) is a unique identifier that tracks the end-to-end process. Activity is the activity name, and timestamp is the activity end time.

There are many tools available to analyse this data, but for Python developers I recommend pm4py. It’s easy to set up: take data from logs, format it, save it, read it with pandas—and you’re ready to go.

There are many possible visualisations, but the Directly-Follows Graph (DFG) is very beginner-friendly and gives process insights quickly. A frequency graph shows case counts, and a performance graph shows time durations.

Now, the most important question: how do we find a bottleneck? A performance graph shows the average time spent in each activity. If an activity stays open for a long period, that’s a cue to analyse it further—it may be an exception path or a process where resources aren’t trained well. The action item would be to improve that activity. It can also reveal automation opportunities or GenAI use cases.

I have added a project related to this—please check the project section antahai mining project and share your comments.