|
How to Perform Root Cause Analysis on Custom Process Anomalies |
Scroll Previous Topic Top Next Topic More |
Probably, the default process behaviors that are provided by the MPM Root Cause Analysis Template do not cover every anomaly in a specific process. Therefore, it is possible to aggregate custom process anomalies to the root cause analysis.
First of all, we need to specify some guideline to create a new problem.
•Problems need to be created within a table (existing or new) which is connected to the MPM Data Model by either the key field "CaseID" or the key field "ProcessPathUniqueNo".
•The anomaly needs to be created as a binary flag within the data model to be recognized correctly by the RCA. The value "1" of this flag means that the case or process variant shows a problem while the value "o" means that the case or process variant shows no problem.
•It is important, that the binary fields get a unique and readable name, because this field names will be displayed as the anomaly descriptions in the front end.
|
The anomaly has to be created for a CaseID or a ProcessPathUniqueNo as a binary flag, where the value "1" represents the problem and the value "0" represents that there is no problem. Give the flag a readable and unique name. |
To create a new anomaly for the root cause analysis the following steps are recommended:
1.Define the anomaly and think about how it can be detected in the data.
2."Simulate" the calculation of the anomaly in the front end to understand which fields of the data model are included and which fields need to be calculated in the script.
3.Create a new tab in the script editor, place it before the tab "5. Root Cause Analysis" and give it a name, for example "Custom RCA-Rules".
4.Calculate the new flag and connect it either to the CaseID or the ProcessPathUniqueNo.
|
Be cautious not to duplicate table entries, revise this in the data model, if the table with the new flag has more CaseID than the CaseTimes-table or more ProcessPathUniqueNo than the pa_activity_log, something went wrong. |
5.Write the new flag into the inline table "RootCauseAnalysisFlags" that was introduced in the section Use Root Cause Analysis on Your Data.
In the example below the flag "My new process anomaly" was added in the first line.
RootCauseAnalysisFlags:
load * inline
[CalculateInRCA, |
FlagName, |
FlagTable, |
ObservationType |
y, |
My new process anomaly, |
CaseInformation, |
CaseID |
y, |
Happy Path Fitness Below 80%, |
ProcessPathConformance, |
ProcessPathUniqueNo |
y, |
Case Lead Time Exceeds 75 Percentile, |
CaseTimes_$(mvProcessDescription), |
CaseID |
y, |
Case Lead Time Below 25 Percentile, |
CaseTimes_$(mvProcessDescription), |
CaseID |
n, |
Total Case Event Duration Exceeds 75 Percentile, |
CaseTimes_$(mvProcessDescription), |
CaseID |
n, |
Total Case Event Duration Below 25 Percentile, |
CaseTimes_$(mvProcessDescription), |
CaseID |
n, |
Total Case Idle Time Exceeds 75 Percentile, |
CaseTimes_$(mvProcessDescription), |
CaseID |
n, |
Total Case Idle Time Below 25 Percentile, |
CaseTimes_$(mvProcessDescription), |
CaseID |
n, |
Case Lead Time Exceeds Median, |
CaseTimes_P2P, |
CaseID |
n, |
Case Lead Time Exceeds Average, |
CaseTimes_P2P, |
CaseID |
]; |
|
|
|
6.Run the data load. The new problem will be investigated and the results will automatically show up in the front end.
If any problem with the custom anomaly should occur, think again: is the anomaly a flag with the values "1" or "0" for a process variant or case?
