A few weeks ago I needed to visit a Dutch hospital. I had an appointment at 14.30, nothing serious, just a check. Arriving in the waiting room at 14.20 I had to pick a number. At that moment I turned from André Helderman into number 452. There were 3 desks. On a screen I saw number 452 had to go to desk 1 to register, so intuitively I walked to the left. Then I looked up and saw this:
Stupid me, obviously the desks are numbered from an employee perspective, not from a client perspective. I corrected myself and walked to the desk on the right side. A pretty strange customer journey, already during the first seconds.
Of course the waiting room is not the core business for a hospital, getting proper medical treatment is much more important, but it does make clear that this hospital organization is thinking inside out instead of outside in.
In the meantime it was 14.30 so I used some time to think about business process modelling. You have to do something, right?
If I would model the client process of getting a hospital check, it would start with something like a reference from a family doctor (that is some kind of a portal towards the hospital, we have in Holland), the making of a first appointment to see the specialist, a second appointment for example to have some blood taken and a third appointment to discuss the results. Something like that.
In a classical process model this would look like a pretty (c)lean and straightforward process.
Process modelling based on process mining
14.40: I was thinking: what if I would use process mining to discover how the current process actually goes? Process mining is based on event logs. Every registered action, including timestamp, is used to make the actual process visible. Doing that it would become visible what the waiting time is between registration and entering the doctor’s office. Averages, minimum, maximum, per day, per hour of the day etc etc. I would be able to make it all visible in patterns and draw smart conclusion. Also shortcuts in the process (so called elephant paths) would be discovered and the amount of cases for which the process is aborted, for example by patients leaving before it is their turn and on which moment in the process this happens. It would learn us much more than the clean theoretical process models.
For more info on process mining: https://www.coursera.org/course/procmin
Process mining based on Internet of Things
In the meantime it was 14.55 and people around me were getting pretty annoyed. I wasn’t; I didn’t mind a bit more time to think further: what if we would use Internet of Things concepts to get even more data in the event logs? For example by having the doors in the hospital censored or even a step further: have the patients connected via a bracelet or a chip. Andreas Sjöström did an interesting experiment, boarding a flight with an NFC implant:
Walking patterns through the building would be visible for analysis including the few wrong steps people take when they approach desk 3 instead of 1. I am pretty sure the numbering of the desks would be corrected pretty fast. Process mining is a very useful approach to improve client processes.
15.10; still waiting. The moral of this story: undesired waiting time can lead to useful insights.