Semantic Checks and how to track them

Hi Peter,

I cannot answer your questions but I have some general remarks. Semantic checks can be useful for a modeler not to oversee any syntactical issue or may be given layout or mandatory data demands added company specific to the checking program. But that's it. The quality and correctnes of a model can only be very basically checked by a script. The term "semantic" can lead to the conclusion that there are any content related checks. This is not the truth. Whether a model is a correct representation of an actual process or a desired target process can only be checked by discussions in a team of people having a deep knowledge about the business starting from the companies strategy. So any automated approval is one little step but not the solution to end with good and helpful models.

Regards

Klemens

There are different possible strategies for how you can manage the models' quality. It all depends on how you define your governance and QA process. I can just share my common strategy that works for us the best. Of course, we use scripts for this purpose, but standard scripts, not semantic scripts. The idea is to get the list of issues from multiple processes in Excel simultaneously. So the scenario is the next: the QA selects the necessary process on the VACD diagram (it can be one or multiple processes on L1 or a specific process on L3, and then the script will go down by levels - L2, L3, then L4 Process level, and checks all L4 diagrams, EPC or BPMN, depends on client). For the project/client we define the Check categories (Error types column on the picture), and then for every category we define rules (Description of errors). If errors are identified, we can see the number of these errors in the column and descriptions of these errors. 

The potential scenario could be the next: modeler reported that the batch of processes is ready (sometimes we use another report that tracks the process statuses, and i,f e.g. the status is changed to "Ready for QA", it is time to start the Semantic checks), then QA starts the report from the selected e.g. L3 and gets the Excel report for all L4 EPCs for that L3. After, if there are errors, it also depends on your QA process: or you return these processes to the modeler, or QA fixes them himself... all depends)

 

Hello Peter,

I re-wrote the standard macro for mandatory semchecks into a report script in order to run it from an approval workflow as a quality gate before creating review tasks for approvers.

Please send me an email to discuss it further. I am afraid I cannot offer just the code because it may not be suitable out of our context.

Regards, Martin

Dear All,

thanks a bunch for your replies. As we are in a very small team supervising a global team of external process modelers, it is crucial for us to know exactly what they are modelling. The input provided by the processes are key input for blueprints and requirements engineering.

We could manage to find an API call to trigger the standard semcheck in ARIS. Very helpful as I want to keep a central point for managing rules. Other than that we continue to export all meta data into a central DB which is used to feed Tableau flows to generate data quality dashboards. That works amazingly well. I can only recommend to explore this.

Have a good start into the week

Peter