The process is never the process

Organisations need to be increasingly adaptive to respond to changes in their environment. An important aspect here is the ability to flexibilise and accelerate processes. To know where processes need to be adapted, it is important to know processes. This does not seem that complicated. But do you know how your processes really work?

Many organisations think they do. Process designs have been made and, based on these, an automated system has often been set up to support the processes. But do the processes also run as the organisation has designed them? And how often do processes go right the first time? As an organisation, do you have insight into how often, for instance, rework or work arounds are applied? In our practice, this insight is often lacking. The conceived process is almost never the real process. We are then talking about the organisation's hidden factory. Nowadays, systems can help and show you what really happens in your organisations using process mining.

Organisations provide products or services to its customers. To deliver products and services, organisations use processes. All activities performed between customer demand and actual delivery are called a process. To improve processes, it is common to start by understanding the current situation. This is why we record processes. To do this, we organise workshops with subject matter experts who tell us how they think the processes are running now. In some cases, customer requests are followed from start to finish in order to map the flow. The problem is often that these workshops, however comprehensive, do not tell the whole or the real story. We talk about a hidden factory created by extra actions, unconscious choices, controls, waiting times or other actions that are considered exceptions. Most waste in processes is often precisely in this hidden factory.

What is process mining?

Process mining is a specialism within data mining. Data mining is used to discover patterns in existing data. It is also used to predict trends in the data. Process mining focuses on the data within processes. By extracting data from IT systems, process mining gives an exact picture of how a particular process functions. In turn, one uses this information to improve processes. Data mining thus focuses on data in general and process mining focuses on events (events and processes).

(Grigorova, Malysheva, & Bobrovskiy, 2017)

In practice, process mining fills the gap between data mining and process analysis by pinpointing exactly where the bottlenecks are. With process mining, process models are discovered based on data. Literally tracking and analysing the so-called event logs provides detailed insight into the processes, revealing bottlenecks and enabling highly targeted optimisation of processes. In addition, process mining provides insight into where bottlenecks have arisen where data mining only indicates certain trends.

(van der Aalst, 2016)

In practice:

One major mortgage lender has figured out its process in the steps below:

1. Registering the application
2. Processing or modifying the application
3. Upload documents
4. Checking documents
5. Issuing interest rate offer
6. Send final interest offer
7. Loan closing
8. Implementing loan in administration

The organisation struggles with excessively long lead times and consequently dissatisfied customers. Although the processes were designed to have a maximum turnaround time of 10 days, in practice little of this appears to happen. Staff indicate that the process is broadly followed. Admittedly, there are occasional cases that require some extra work because the information is incomplete, but they feel this is only a small percentage. To help analyse, the organisation decides to apply process mining.

First, a discovery method is applied. This method makes the process transparent on the basis of the data. Through so-called event logs, the process is recorded. Event logs are events that are recorded and stored. In this way, every action performed by the system or the employee gets its own record. For each case, it can be made clear which events have taken place for that case. If you do this with a large quantity of cases, you get great insight into the process with the quantities that follow a certain course. Thus, the actual process is displayed. So a process is mapped purely on the basis of event logs. There is no comparison with the conceived process at that point. In the case of the mortgage consultant, it is striking that, out of 100 cases, as many as 45 go from activity 4 checking documents back to activity 2 Processing or adjusting the application. Apparently, incomplete or incorrect information is often supplied. It is also noticeable that after activity 5 Issuing interest offers, processing times are very high in 15 of the 100 cases. The processing time between activity 6, 7 and 8 varies from 8 days to 24 days.

The second step is conformance checking. This involves subjecting the process previously established in the discovery phase to comparison with the actual process. The difference between the apparently perfect process and reality is brought into focus. For example, a process may be conceived as ACBD but in practice run as ADBC or follow several different routings.

So the outcome of conformance checking is the difference between the intended and actual process model. The observations made earlier for the mortgage lender show that there are some differences in workflows and differences in predicted turnaround time. Possible substantive outcomes are that process steps differ in order or that some products go through the process differently than expected.

In the final stage, enhancement, information from the event logs is used to make process improvements. The event logs contain information showing, for example, peculiarities in turnaround times or percentage differences in approval or non-approval. The follow-up step is to check whether the actual process is desired or needs to be adjusted to the intended process. In addition, a follow-up step may be to examine what kind of products follow a different routing and for what reason. The mortgage lender finds that information is often delivered incorrectly. They decide to better inform their customers by clarifying the online application for applications. In addition, they are improving control over the approval process.

Use of process mining

Process mining can be used to make processes more efficient. Insight into event logs identifies bottlenecks that drive inefficiency. In addition, process mining can also generate savings in both time and money.

Process mining can offer an organisation insight into, for example, operations that are performed in very different ways or process activities that are not performed at all. This gives the organisation tools to better assess processes and improve accordingly.

For any process supported by a system that does event logging, an organisation can decide to mine. All process steps become visible. Specifically, this means that when there are different routings for the same or a different product, this is made known through process mining. In addition, it becomes clear how long an activity takes and why a process might be less efficient. The path each product or service or operation takes is made visible.

Of course, there are a number of preconditions that need to be met to apply process mining properly. These include data quality and the ability to retrieve event logs. In addition, (the same) system often does not support all activities. This makes mapping the actual process a lot more complex. In a subsequent article, we will discuss these preconditions and how they can be dealt with.

Authors:

André van Hofwegen, Lisa van den Berg, Marc Hählen and Rick Moolenaar (Consultants at Improven)