Process Mining Data Essentials

Raw data is only the starting point. Process Mining relies on clean, well-structured event data. Learn how to assess, prepare, and improve your data for reliable process insights.

Chapter 1: What data you need to get started

Imagine you run a lemonade stand but have a terrible memory. To understand how well your stand is doing, you decide to track three basic details:

• Customer ID (CaseID): This number identifies each customer. It shows whether the same person comes back for more lemonade or to complain about a sour batch.
• Action Taken (Activity): This records what happened. Did you “Take Order,” “Prepare Lemonade,” or “Resolve Angry Customer Complaint”? Hopefully, not too often.
• Action Time (Timestamp): This records when each action happened. You need the sequence of actions to understand the process.

With these three data points, Process Mining acts like a small detective at your stand. It shows the basic customer flow, identifies bottlenecks such as slow lemonade preparation, and reveals whether some customers complain more often than others. That gives you a clear starting point for improvement.

Here’s an example of what the data might look like in a table:*

This may seem like very little information, but it gives Process Mining enough to start asking questions and uncovering basic insights into your lemonade stand’s efficiency.

Chapter 2: The Case of the Missing Lemonade Logs

Our lemonade stand was a smash hit. Customers loved our secret recipe, mostly, and business was booming. But success brought a new challenge: we were overwhelmed by customers. Lines were long, tempers were flaring, and worst of all, we had no idea why.

Remember the small detective we hired, Process Mining? It couldn’t work miracles. It needed reliable information, and all we had were a few scribbles on a napkin. Here’s where things got messy:

Cleaning up this data mess became a new project. In the next chapter, you’ll see how a little detective work and help from our data-focused Process Mining solution helped us optimize the lemonade stand and become the envy of the neighborhood.

Chapter 3: The Great Data Dig

Our lemonade stand was a smash hit, but the lines were a nightmare. We knew Process Mining could help, but first we had to provide reliable process mining data. That meant a deep dive into data extraction: finding hidden clues about our customers and turning them into information Process Mining could interpret.

Here’s what we discovered:

• Treasure Hunt: Sometimes, the data was like buried treasure, hidden in dusty corners of our systems, including web pages, emails, and PDFs. We had to become data archaeologists, digging through old files and using tools such as screen scraping to find the information we needed.
• Lost in Translation: Even after we found the data, it was not always clear. Some clues were scribbled on napkins, which made them unstructured data, while others lacked metadata. We needed data standardization to interpret it all.
• Focus is key: With thousands of tables across our process mining data sources, it was tempting to collect everything. But just as you would not try every flavor combination at an ice cream shop, we needed to focus on the questions we wanted to answer. Did customers take too long to order, or were we the bottleneck when making lemonade? These questions helped us prioritize which data to extract.

It took effort, but with curiosity and careful work, we uncovered a valuable set of process mining data. In the next chapter, we’ll see how we cleaned it up and got Process Mining working for us.

Chapter 4: The Data Detox

We had a mountain of data after our extraction efforts in Chapter 3. But the dataset was inconsistent: some customer information was useful, some entries were random notes, and much of it was irrelevant. It was time for a data detox.

Filtering became our new best friend. Think of it as sorting through a messy toolbox. We started with broad, coarse-grained scoping during extraction. Now it was time to apply detailed, fine-grained scoping.

Here’s how we tackled the filtering challenge:

With the data mostly clean, it was time to apply Process Mining in the next chapter. We would explore discovery, conformance, and enhancement to diagnose problems at the lemonade stand and improve its efficiency.

Chapter 5: The Data Makeover

Our data detox in Chapter 4 helped, but one crucial step remained before we could apply Process Mining: the data makeover. Imagine a customer arriving with a crumpled dollar bill. You would still accept it, but a crisp, clean bill would be easier to handle. That’s the idea behind data cleaning.

Here’s what we needed to do:

• Case Closed: A process is like a customer’s journey: it has a beginning, middle, and end. We needed to connect every event related to a single customer case, including the order, wait time, and final lemonade delivery. Think of it as organizing all receipts from one customer visit.
• Speaking Process: Our data did not always describe the process clearly. Activities needed to represent defined status changes in each customer’s case. For example, “Customer Happy!” was too vague. We needed a clear status such as “Lemonade Delivered.”

It was not the most glamorous part of the project, but careful data preparation and clear definitions gave us a clean dataset. With Process Mining working on this transformed data, we could uncover the causes of long lines and turn our lemonade stand into a more efficient operation.