Process modelling versus process mining
Two ways to improve processes: employee-led models, and quantitative data analysis
Most organisations use one of two methods to find bottlenecks and improve their processes.
The traditional way – made popular through methodologies like Lean – is model-based process improvement. This uses information from employees and stakeholders to build a model of the current process. This process model is used in workshops and interviews, and the resulting improvement points are built into a new process design.
But an alternative, data-driven approach is becoming increasingly popular. Process mining uses specialist software tools to retrieve and analyse data that’s naturally captured in your information systems during the process, such as messages, transaction records, and event logs.
Instead of building a model based on the subjective views of stakeholders, process mining uses these quantitative data sources to visualise what employees actually do. Because this is objective, it can reveal unnoticed bottlenecks in the process, like re-work, lead time overruns, and hidden deviations from the process path.
This quantified, objective approach introduces some new possibilities that are difficult with a process model approach. It gives a basis to objectively compare similar processes, for example at parallel sites. It also lets you establish a baseline measurement, so you can quantify the impact of your process improvement work.
Comparing the two approaches
Process modelling and process mining both have strengths and drawbacks
Both approaches to process improvement have clear strengths to recommend them. But both have drawbacks too. Choosing the right one will depend on the needs of your particular organisation.
The process model approach is relatively quick and easy to achieve, with few barriers before you can begin. And importantly, the workshops naturally get stakeholders, employees, and process owners to engage and participate in the improvement journey. However, its conclusions are subjective. If the process model you create is not accurate, the improvement actions you identify may not have the desired result. Similarly, which bottlenecks are most important is a matter of opinion, so you may not focus on the areas with the biggest impact. As an objective, data-led approach, process mining eliminates these issues; the real process flow is visualised and supported with facts. However, it does present difficulties of its own. Most obviously, there’s a lot of work to do before you can even start. The process mining tool itself needs to be learned, and it an be time-consuming to find, organise and structure all the right data streams so they’re suitable for analysis.
You might find there’s a lot of irrelevant data – for example, if your process involves a lot of messaging applications. Conversely, if some parts of your process aren’t tracked in an information system, you may find there are areas that produce no data, and therefore can’t be mapped.
Combining approaches and avoiding pitfalls
One further alternative would be to use the two approaches together, and combine the advantages of both.
For example, you could use process modelling techniques to build momentum, involve stakeholders, and identify the main bottlenecks. You could then verify this against the analysis from process mining to complete the picture, establish priorities, and set baselines to measure your improvement. From here, you could go a step further and use process simulation to see how the improved process would perform. The three aspects might work together as follows:
• Process mining: What do we have now, and where are the bottlenecks?
• Process modelling: What would we like the process to be?
• Process simulation: How would this work in practice?
Closing the loop in this way helps to rule out one of the potential pitfalls of process improvement: causing unintended side effects elsewhere. It’s important that the picture includes the entire process from end to end, seeing how the proposed improvements to various sub-processes work in relation to each other. There are other possible pitfalls that, handled well, can become success factors as you improve your process.
First, there is a danger of expecting process mining tools to automatically deliver insights by themselves. In fact, the information needs to be examined carefully to see what business value it offers – whether it’s improving efficiency and effectiveness, or reducing risk. It’s also important to choose improvements carefully, to build momentum within your organisation. Based on the data alone, you could be tempted to change the primary process – when in fact you might achieve success faster by working on a process with a defined beginning and end, and clear workflow support.
Another temptation is to try and answer every question at once. Your initial analysis will often raise further questions, so it’s important to separate the issues clearly based on the business value you want to achieve, and tackle them individually. One good way to address this is to propose a hypothesis for each question you want to answer, and use process mining to prove or disprove them in turn. Remember, however, to let the facts speak for themselves; don’t fall into the trap of trying to justify a subjective opinion.
Getting started is becoming easier
It is now easier than ever to use process mining techniques to find bottlenecks and supplement, quantify, and shape your improvement work. Obtaining the specialist knowledge and software were once a significant barrier – but this is no longer the case. As the market has matured, process mining software has become easier to use.
Meanwhile, there are exciting streams of process data being logged. It’s now possible to harness barcode scanning, or take feeds from wearable technology and the Internet of Things. The field is still developing. But given these changes – and the advantages it offers – we expect the popularity of process mining will continue to grow for years to come. If you’d like some help getting started, our data engineering experts would be happy to chat.