So, I'm working with another great customer happily process mining their sn_customerservice_case CSM table when I notice they have empty configuration items device types. This draws my attention as the time to go from pending-> resolved for these cases with empty configuration item device types is 5days (figure 1), which seems like a long time.
(figure1 - pending to resolved taking 5 days)
So I think this a great place to through cluster analysis at the data. I right click on the 5d/23.1k node and right click to cluster. But then I get an error about "no process configuration for the sn_customer_service_case table (figure 2)? So what gives?
(figure 2 - error when clustering the sn_customerservice_case table).
Turns out in the Rome instance I am working on I need to create a cluster solution definition against the sn_customerservice_case table and then associate that solution definition to the process configuration for that table.
At a high level here are my steps that I used to resolve the issue:
1. All > Predictive Intelligence > Solution Definitions > Create Clustering Solution Definition
2. Create your word corpus and cluster definition. If you are new to setting up a cluster, no problem. It's easy, for a step by step go to NowLearning and search on this excellent PI primer "K21 Accelerate Incident Resolution for Agents using AI/ML"
3. No need to train your cluster solution.
4. Navigate to All > Process Optimization > Process Configurations> Process Configurations list.
5. Select New.
6.Select the table to associate a clustering solution definition to.
7. From the Clustering Solution Definition field, select the search icon.
8. From the Clustering Definitions list, select an existing definition from the list
9. Select Submit
Here's the ServiceNow doc for reference.
Now you can go back, right click on your Process Optimization node where you want to run the cluster and ckick clustering. When it's done clustering you will see "view clusters" (Figure 3).
(Figure 3 -view clusters)
Clicking into View Clusters shows me the cluster concept, which are the top ten words common to the cases in each cluster. This is useful because it helps me answer my original question, "what types of cases have empty configuration case types?" (Figure 4)
(Figure 4 - Biggest clusters and click view all clusters in Treemap).
Pro Tip - if you don't have a ton of data you can go into All > sys_properties.list. Search for *cluster. You will see min and max row settings, you can reduce your min setting down to 100.
HTH and Happy Process Mining!
