Retrospective of a Process Mining Project

We recently finished the first phase of a Process Mining project that has taken us almost a year. In the project, 10 processes were planned to be analyzed, 11 actually were, and finally in 9 we reached a complete analysis. 

The project was not only about analyzing processes, it was also about introducing the concept, "evangelizing" key management people and defining and implementing a methodology that would make it easier to integrate process improvement orientation into the organization's practices.

We consider it a good idea to do a "retrospective" when we finish a project (in this case a phase) and review how it went, what should have been done differently, what decisions should have been different, etc. Everything, of course, with the purpose of gaining knowledge and that the next project goes better.

It is now time to do the retrospective of the above-mentioned project and we thought it would be interesting to share it here (not everything but at least a good part can be told).

Here we go:

Achievements:

1. Diffusion and Evangelization. 

An important objective of the project was to demonstrate the value that the application of Process Mining could bring to the organization's process improvement initiatives. To do this, it was important to explain the concept and convince the decision makers and influencers within the organization. We approached it in two steps. An initial step with workshops to explain the concept, application and benefits of Process Mining, and a second where for each process we showed how the results could be used to improve the process. Not to everyone but we were able to convince a large majority.

2. Implementation of methodology and practice

Based on our own methodology, we were able to adapt it to the specific case and detail required by our client. It may seem strange but we can proudly say that the client will be able to continue on his own even if we are not there - if he dedicates the needed resources of course.

3. Image of the project's sponsor as a valuable business contributor.

It was not a sought result but we managed to strengthen the image of our sponsor in the organization as a generator of innovation and contributor of value solutions to the business. It was very gratifying to witness how the business areas congratulated our sponsor when we finished the workshops to share results and design an action plan.

4. "Actionable" results in most of the processes analyzed.

In almost all the processes we reached conclusions that allowed us to define a plan of specific improvement actions. And very important: most of these actions came from the business areas themselves.

5. Identification of deficiencies in how transactions were recorded in the systems. Another not sought result, but we identified some cases where the information of the process was not being correctly registered. This was very relevant due to regulatory requirements of traceability.

Lessons learned

1. Involve more and better the technical team that extracts the logs from the systems used by the processes.

We must dedicate more time to them, give them more context, not only the format of the file and the basic ideas. You have to explain them the purpose and even share with them the problems, not only the technical ones, that you encounter when you are preparing the log to make it viable. You will save time and get a higher quality log. We learned it a bit late and that meant a wear and tear that could have been reduced. 

2. Add the analytical business dimensions to the process metrics model.

From the very beginning you have to be thinking about how you are going to measure the process from a business perspective. You have to try to fit the well-known metrics that are valid for any process (times, queues, rework, automation, etc.) into the analytical model that is used by the key holders of the process. For example, if you do a time analysis of a customer service process, you need to be able to segment that analysis by the type of request, the type of customer, the amount associated with the request, etc. This has a direct impact on how you enrich and process the log. This point is important in order to be able to compare performance and compliance between segmentations.

 3. Incorporate economic vision into the analysis.

An analysis is valued differently between business (euros) and analysts (times, amounts) so try to add economic context to any metric. For example, if you find that 2% of payments do not respect the segregation of duties, you must also provide the value in Euros of that 2% of payments. This again has a direct impact on how you enrich and process the log.

4. The presentation of the analysis of a process is different according to the recipient. The "executive" reporting of the project is presented in a different way than the "operational" one. The results of the analysis of the order-to-cash process should be shown to the CFO in a different way than they are shown to the business analyst of the process or the individual who is making the collection. 

Next challenges:

For the second phase we intend to implement the following improvements:

1. Get better logs:

• Add to the log more information about analytical dimensions in order to obtain better metrics and the possibility of benchmarking.
• Quantify (value and/or cost) in each case of the log. Not in all processes will be direct.
• Add events that occur in complementary systems to the main one.

2. Extend the practice to more systems.

In the first phase we worked with three IT systems (one application in a more or less standard BPM and two custom made applications). In the second phase we want to add more systems, which will require a re-adaptation of the methodology.

3. Standardize and automate, at least in part, the analysis presentation reports.

In the first phase it was a very manual part and we believe that it can be improved a lot. We'll be using an enhanced BI tool for that.

4. Search for a more effective and complete way to translate the analysis results into improvement actions.

We had been doing it in a somewhat spontaneous way in the workshops of presentation of results. We have to find a way to do it with a workshop dynamic that generates more and better ideas. 

5. Incorporate the possibility of comparing the same process in two different moments of time.

This will be very important to evaluate the result of the actions implemented in the first phase.

We hope it has been of interest. Comments are very welcome.