Saturday, April 12, 2008

New BPMS feature: process mining

Recently I attended a meeting where a BPMS vendor (Pallas Athena, presented a new (for me) feature: process mining.
They worked on this concept together with the Eindhoven University (Van der Aalst, a well-known name in BPM research, also see and

Basically, the concept is:
- the BPMS reads all kinds of logfiles from various computersystems.
- The BPMS analyses the logfiles and will construct a processmodel automatically
- The data + model can be used for animation and simulation

That's interesting for multiple reasons:
- It usually takes quite some time to construct a procesmodel manually - talks to stakeholders, workshops, etc. Process mining can save time (and cost)
- Even when interviewing stakeholders, they might not have the correct overview and the modelled process might not match reality. Process mining can confront stakeholders and improve model quality.

There is catch of course:
1. The logfiles need to provide data that a BPMS can interpret - aka some type of task identification, a timestamp and a unique ID (of a certain flowing object, such as customer, order, process, etc)
2. Process models that are "mined" will only contain the reality as is presented in the logfiles, and no more. They will miss:
- Business rules for decisions and assignments (why did this instance go left, why was person X the performer of the task)
- Manual activities or activities not traced in logfiles

Still, I saw some cool stuff, that could have added value in BPM efforts:
1. A logfile of a helpdesk application was analysed. Step by step you saw the processmodel appear and get refined, based on the logfile entries

2. A time-based animation could be run, based on the original logfile, showing the flow (little dots) through the process model. This was quite interesting, because it visually showed the execution, quickly pointing to bottlenecks and opportunities for improvement
(Note: if would even be better if the system would be able to present possible improvement actions, based on process improvement frameworks such as Lean and Six Sigma)

3. When the animation was shown to the stakeholders, they did not believe the process - certain flows where against company policy, but did happen. Compliance issues became clear.

4. Based on the logfiles, various KPI's could be calculated. In addition, various statistical facts could be derived, usefull for subsequent simulation

5. If the eventlog also contained userdata, it was possible to analyse and visualize the "social map" - who is involved when, where and is linked to other people.

So, even though the technology is quite new, needs refinement and has limitations, I do see interesting benefits:
- Quickly "mine" a starting point for a processmodel, and then add tasks and business rules
- Visualize currect processes and analyse bottlenecks

For more information - see:

Disclosure: I have no commercial relation with Pallas Athena.

No comments: