It has been great to participate in Process Mining projects. Typically in them we make a lot of discoveries based on the transaction-level data. Some of the typical discoveries include longer than expected lead times, skipping of required activities, as well as large amount of extra activities & rework. While it is great to make these discoveries, the operational business people typically need more detailed information about the root causes in order to start corrective actions and business improvement projects.
Root causes can be identified with a traditional interview method by asking a lot of questions from many people. However, there is a better way. Using QPR ProcessAnalyzer’s influence analysis one can use data mining methodologies directly integrated with the process mining discovery tool. In practice this means that a human user points out a discovery by identifying group of process cases in QPR ProcessAnalyzer user interface and then clicks the “Influence Analysis” button. QPR ProcessAnalyzer will run through the whole dataset and find out how the selected group of cases is different from the other cases. These results point out the most important causes for the original discovery.
In practice many organizations are nowadays using QPR ProcessAnalyzer Influence Analysis as a regular part of their process development cycle. Typically on monthly basis the root causes for identified problems are re-evaluated and performance data is used by operative business people to monitor and manage their business improvement activities. These business review root cause analysis results are shown in benchmark format pointing out the worst-performing segments as well as the best-performing segments.
The picture below is an example of an Influence Analysis. This analysis shows root causes for process cases having a total duration longer that 10 weeks. On average 20% of cases (2.129 cases out of 10.491 cases) take more than 10 weeks. The most probable root cause is the Dallas Region, since in that segment 54% of cases run late. As a benchmark we also see that New York is doing very well in this example, only 11% of cases run late. Influence Analysis is thus able to find the most relevant attributes (in this case Region) and then pinpoint the most relevant attribute values (here Dallas has very long durations and New York is very fast). The analysis also shows that there are other case attributes that correlate with total case duration, for example Customer Group Kids cases take clearly more time than Customer Group men.
Sometimes there is no clear root cause for a certain business process problem. Then we typically have two possibilities: A) the analysis data does not contain any indicator of B) the process is failing with equal probability in all segments. In the latter case it may be a good idea to improve the whole process using for example traditional process improvement and re-engineering methods.
Summary: whenever you want to improve your business process, it is a good idea to use process mining to discover actual process problems and Influence Analysis to find out the root causes for problems
Contact us to learn more!