Simulation - Shared best practices and learnings

Mark Barnett on August 31, 2009

Hello group,

I have been using ARIS Simulation capability for several months now and am interested in sharing best practices and learnings around the use of this capability.

In my company simulation has not been used heavily, however the work that I am doing is showing promise of leading to the use of simulation as a standard tool in support of capacity management, forecasting, process improvement, and so on. I have had success modeling a core end-to-end operational process in the financial services industry, which is supporting a merger between two large companies which are moving towards a single platform for their core process of retail loan origination.

One of the areas that I have been looking at is optimizing the head count ratios of the various roles involved in this process, as well as exploring the dynamic response of the process to spikes in incoming loan application volume, and how to mitigate that response in the most cost-effective manner, e.g., using overtime.

To get the ball rolling on this discussion, I pose the following questions to others using ARIS simulation:

1) For what problem types are you using the simulation tool, and what level of acceptance from senior management in your company (if applicable) are you receiving? What steps have you taken to gain their acceptance and how successful have those steps been?

2) What have been the biggest challenges you have found in trying to use ARIS simulation and how have you overcome those, if you have, or if not, what steps are you taking in an effort to overcome those challenges?

I would welcome any other thoughts, comments or questions about ARIS simulation from this group!

Thanks,

Mark

Tags: #simulation; #bpm

Notify Moderator

11 Replies

Ivo Velitchkov on August 31, 2009

Hi Mark,

Thank you for bringing up this topic!

1) Capacity planning, location and elimination of bottlenecks and a little bit for optimisation of process instantiation

2) I've recently found out that when choosing "relative" time, schedules for process instances and working time are ignored. All instances are equally distributed according to the frequency of the starting event not taking into account the actual schedule assigned to it. I didn't find it documented so I wonder is it intended behaviour or a bug. With "absolute" time everything works fine.

I like the new simulation experiment capability.

Regards

Ivo

Like

Sign in
or register to reply.

Notify Moderator
MB

Mark Barnett on August 31, 2009

Hi Ivo,

Thanks for your comments. I haven't tried using relative time, so I can't share anything there.

I will tell you one area I shared with IDS Scheer as a key unmet need for me in current ARIS: the ability to output "snapshots" of the process at specified internals, e.g., at the end of every month of a year-long simulation. (FYI, iGrafx offers this).

Another "missing" capability, from my perspective, is the ability to get a "trace" or time history of specified parameters as an output. Constructing this from the data available in the standard output files can be quite difficult, depending on what one wishes to have a trace of.

This said, I like the ARIS Simulation capability overall, and have been able to produce some interesting and value-added results so far. Being able to identify dynamic response of a business process to a disturbance (e.g., a short-lived surge in input volume) has been quite instructive. I know of no other way to obtain such dynamics for business processes on a predictive basis.

Best wishes,

Mark

Like

Sign in
or register to reply.

Notify Moderator
Ivo Velitchkov on August 31, 2009

Mark,

May be with the simulation experiment capability you'll be able to produce some statistics which will give you "snapshots" at given intervals.

What you said about identification of business process responses to disturbances, sounds very interesting. Can you elaborate a bit more on this.

Thank you!

Ivo

Like

Sign in
or register to reply.

Notify Moderator
MB

Mark Barnett on August 31, 2009

Ivo,

If I knew how to feed the outcome of one simulation into the next in an experiment as an "initial condition" that could work, but as far as I know such a thing is not possible in ARIS. What needs to happen - and I believe some work may be going on within IDS to do this in a future release - is for the simulation to "dump" statistics at predefined periods during the simulation.

Regarding response to disturbance, what I am doing is pretty straightforward. I run the simulation with a specified instantiation rate (input volume) for the entire year (assuming a full year simulation), and "overlay" another added instantiation ("surge") on top of that over a period of, say, 1 week. That throws more instances on top of the background, steady-state volume. In my case I double the volume for 1 week out of the year, then return it to normal and watch what happens.

The response is quite strong, and what I see is a dramatic increase in cycle time that relaxes back towards the steady state (non-surge) cycle time, but over a very long period of time. I can take that same simulation and now rerun it with the addition of some specified amount of overtime for one or more roles, and see what it takes (how much overtime, for which roles, and for how long a period of time) to minimize the impact of the surge. Some interesting results have come out of that, such as the need for only one of the three roles in the process to work overtime... not an intuitive result, necessarily, and not how overtime is handled in operations, often, where all roles typically are put on overtime concurrently. I look at this as identification of a low-cost mitigation solution to volume surges.

Mark

Like

Sign in
or register to reply.

Notify Moderator
Ralf Angeli on September 1, 2009 as Reply to Sebastian Stein

Regarding schedules and relative time: Schedules are supposed to be evaluated when using relative time just as in the case of absolute time. For that to be possible the current date and time is used as the start for the relative time period.

The described behavior of evaluating a schedule when using absolute time and not evaluating it when using relative time should not happen and in fact I've never encountered this behavior myself. Would it be possible for you to send a sample database for reproducing this to the ARIS support?

Like

Sign in
or register to reply.

Notify Moderator
Ivo Velitchkov on September 1, 2009

Hi Ralf,

Yes of course I could send a sample DB. Should I send it to the general support mail address or to a specific one?

BR

Ivo

Like

Sign in
or register to reply.

Notify Moderator
Ralf Angeli on September 1, 2009 as Reply to Sebastian Stein

The general one should be fine. You can mention in the mail that the database should be forwarded to me.

Like

Sign in
or register to reply.

Notify Moderator
Ivo Velitchkov on September 4, 2009

Any ideas how to include process instantiation as factor in an experiment? The attribute belongs to calendar event object type which is not included in the object list for factor variation experiment.

Like

Sign in
or register to reply.

Notify Moderator
DJ

Dieter Jansen on October 5, 2009

Hi Mark,

Your "snapshot" idea seems to be the same problem I've come up against on most of the simulation examples I've run, namely the inability to be able to use an ARIS-generated value as an input to the next function. Example: Modelling of a phone queuing system. I want to put in a rule that if it takes longer than a certain time for a call to be picked up, the call is dropped. This would be evaluated at an XOR with appropriate events. However, there is no way to access the randomly-generated time in the previous step. The only way around this is to use some form of statistics, bearing in mind that throughput time values will not be as accurate in this case. Did you say this may be included in a later release? It really isn't anything other than an if-then-else statement, using the result of the previous function. Another example is of an economic order quantity using a general resource object - when the value drops below "n", then trigger a process. I haven't found a way to do this in a straightforward manner. Have you?

I'm also interested to know whether you are using simulation on Business Architect or Toolset. I tend to always migrate back to Toolset because I find it easier to use (or maybe I'm just lazy to learn!)

Regards

Dieter

Like

Sign in
or register to reply.

Notify Moderator
DS

Denis Smirnov on March 22, 2010

Hi Mark!

Now I have task to choose a process simulation tool. The process to be simulated is "multimodal passenger traffic". I have to find out bottlenecks of this process associated with max number of passenger can be carried, max number of required railcars and so on. I've seen some screenshots of ARIS Simulator, but it doesn't meet my requirement to show this to my leadership. I need a graphically customizable tool that could show not only square box of function or event, but show for example railway station (my picture), airport (my picture), trains running between this objects... So I want to know if ARIS Simulator suits me and my tasks or I should look for another suite.

Best wishes,

Dennis

Like

Sign in
or register to reply.

Notify Moderator
Geoff Hook on June 22, 2010

Dennis

why not look at WITNESS, it is a general purpose simulation tool also provided by Lanner which supplies the component integrated into ARIS.

Initially we have plenty of information on our website www.lanner.com and a short video introducing WITNESS to ARIS users is available on ARIS Community http://www.ariscommunity.com/users/geoffhook/2009-12-07-extending-aris-business-simulator

This provides the user with an easy to use simulation environment and full control over logic, data and display including some automatic 3D.

Let me know if you want to take this further

Geoff

Like

Sign in
or register to reply.

Notify Moderator

ARIS User Group DACH meeting