The Simulation feature makes future predictions by simulating the Return on Investment (ROI) before you implement any process improvement initiative, such as Robotic Process Automation (RPA). You can create or manage simulation scenarios from the BPMN feature. You can also access existing scenarios by using the BPA tool.
Every simulation scenario is based on a BPMN diagram. The BPMN can be automatically derived from your data. You can obtain simulation scenarios that are based on your real process from the derived BPMN. You can also create simulations from scratch by starting from the BPA where you can use the BPMN from external sources that integrate with IBM Process Mining.
With the Simulation you can create multiple scenarios by using historical and contextual data from the process. If no simulation scenarios are linked to the project, the panel displays the option to create a new scenario.
The first version of the simulation is automatically set by the application, by appending a suffix “_1” to the simulation name.
When there are simulation scenarios that are already linked to the project, the panel displays the option to create a New simulation or create a new version of an Existing simulation.
If you select Add to existing simulation, you create a new version inside an existing simulation.
After you create a New simulation scenario or choose an Existing scenario, IBM Process Mining automatically generates a pre-calculated Simulation specification based on historical and contextual data of the process.
Here, we are able to make any necessary adjustments to run the simulation:
Simulation version’s name.
Description of the Simulation version.
Number of instances
Number of cases that is generated by the Simulation.
The Simulation generates relevant data that will reflects the distribution of the real process. Underlying correlations are discovered and replicated in the simulation.
Adapt Staff Availability to Full-Time Equivalent (FTE)
Since the "Staff Availability" is gathered from the real process, the measure might be inconsistent when the "Resource" field is not corresponding to a specific resources (but maybe to an office). By enabling this option, when "Staff Availability" is lower than the FTE calculated by IBM Process Mining, the FTE is used as "Staff Availability".
The distribution of the generation of cases, which can be generated by the following Distribution Algorithms:
Base time unit
By default the time is precise to minutes. By selecting seconds, it can be increased to seconds.
Considers the median or the average time of the actual process to set the default of the Simulation specification. According to this choice, IBM Process Mining automatically sets in the simulation all the activities service and waiting times that are retrieved from the data derived model.
Date from which to start the simulated events.
Business hours (Exclude weekends)
The working time in which the service and waiting times are considered. By setting 0-24, you are considering the time referred to the entire day (24 h). By setting, for example, 8-18, you are considering only this timeframe to evaluate service and waiting times. By checking Exclude weekends, you are not considering weekends in the times evaluation.
The fixed Simulation header automatically updates when any adjustments are made to the configuration of the scenario. The Simulation header includes the following actions:
When you made adjustments to the Simulation specification, you can find for each activity a pre-calculated field where you can make any necessary changes to four different groups:
The settings are preconfigured according to the information that is retrieved from the data source, the user can change the information based on the scenario to be created.
In the Settings of each activity, you are able to change different performance indicators to run the simulation:
The activity’s service time must be always greater or equal than its working time: if not, the working time is used also as service time.
To force the instant of the start of an activity based on the user’s configuration. In the Scheduling of each activity, you are able to change different performance indicators to run the simulation:
The activity can be carried out only during specific hours.
Sets a calendar frequency.
Sets a scheduled frequency.
In the RPA of each activity, you are able to configure the data from the activity in the case that it is carried out automatically through a robot. You can change the following performance indicators to run the simulation:
The percentage of the activity that will be managed by robots.
The business hours of the robots.
Number of robots
The number of robots that will work in parallel with the staff to run the activity.
Service time (of the robot)
The service time of the robots. The service time of a robot is equal to its working time, as no interruptions must occur in the robotic activities.
If an automatic activity attribute is configured in the project settings, the RPA default values (excluding business hours) are automatically retrieved from the data source.
In the RPA menu, you can evaluate the impact of RPA on your process costs and performance. For example, you can increase the "robotic quote" of a critical manual activity and simulate how your process would improve.
The "Waiting time" is the time that is elapsed between the end of the last activity and the beginning of the current activity.
In the "Waiting times" of each activity, you are able to change the waiting time of both manual and robotic tasks for any incoming activities to run the simulation. The default values are automatically retrieved from the data source.
Tip: the waiting time inserted here is a lower bound, therefore it ALWAYS elapses in the simulation, not depending on the queues created within the simulated scenario.
Take a look at this article to have a better understanding of waiting time and its components.
Consider the following best practices:
If you need to configure/change the waiting time for a certain activity, try to estimate what is its average transit time (i.e. waiting time that will regularly occur, even if no queue is generated).
When a waiting time is caused by the scheduling of the activity, set the correct scheduling in the related settings
For each Gateway, IBM Process Mining automatically generates a pre-calculated Simulation specification based on all the historical and contextual data of the process. The specification can include the one or more Conditions that are the decision rules that are discovered before you create the new scenario or link to a decision table while you edit the BPMN model.
You can edit the following parameters to run the simulation:
IBM Process Mining simplifies the Simulation configuration by automatically revealing configuration issues that might affect the Simulation results.
|Waiting time less than working time: working time is used as duration||The working time must be revised because the data source detected a minor waiting time.|
|Some activities might have configuration issues||Issue relative to the settings of the related activity.|
|Processing time less than working time: working is used as duration||The configured working time must be revised because the data source detected a minor service time.|
|Variants disabled because manual changes were made in the scenario||If the BPMN was modified manually, it is not possible to reproduce the variants of the original process in the current simulation scenario.|
|FTE not available because Working Time is not defined||The working time of the activity is not defined.|
|Total percentage is not 100||The sum of the percentages of the possible outputs of a gateway is not equal to 100. It is possible to save the scenario but you can't proceed before resolving the issue.|
|FTE not available because frequency is not defined||The activity is not present in the actual process data source. Therefore, the FTE is not available.|
|Resource-related cost is missing in observed data. The estimated cost does not include this cost component.||The cost of the resource is not defined in the process settings.|
By selecting "Run Simulation" IBM Process Mining runs the Simulation according to the changes made.
After you run the simulation, you can see the comparison between the real process and the simulated scenario in Process Diff.
By clicking "Import", you can import the simulated scenario as a new data source in your Process Mining project.
By clicking Versions Compare, you can compare different simulation scenarios
The first scenario that is selected is considered the “before” scenario. The second scenario is considered the “after” scenario.
Changes might occur in:
The type of change can be:
By selecting one specification, activity or gateway, you can visualize the relative list of specific changes.
By clicking Compare BPMN, you can compare the two BPMN models.
A list of all the changes can be displayed.
Consider a bank account closure process: when a customer of the bank requests the closure of an account, the bank must take a series of actions to ensure that the closure takes place correctly.
As this process is human-based, we want to evaluate the benefit of automating one of its critical activities. We can evaluate this process with IBM Process Mining by starting from the real process data, simulating the automation and finally comparing the real As-Is process with the simulated To-Be process.
By starting from the IBM Process Mining model, we can detect the main KPI-critical activity:
We can thus decide to automate this activity: thanks to the Simulation, we can evaluate the impact of automation on our process BEFORE implementing the automation.
We switch to the process BPMN, automatically generated by IBM Process Mining, and click Simulation:
In the Simulation specification, we set a historical arrival distribution and we specify the actual working hours by inserting “8-18” and excluding weekends: with these settings, we obtain a simulation very close to reality.
We focus on BO Service Closure. All settings are automatically retrieved from real data of the process. We reduce staff availability (that is, resource allocation) in the range 3 - 1.
We turn on the RPA settings:
No fixed waiting times are expected in this process, that is, waiting times depend only by the queues that are generated within the process → we must set all waiting times to 0.
We can now run the simulation.
After the run, the average lead time decreased from 17d to 12d and the average case cost also decreased which means that our automation hypothesis can indeed generate benefits within our process.
We can also see that “BO Service Closure” is now no more a KPI critical activity.