Best Practices for BMC Helix ITSM Insights
We recommend that you review best practices for the BMC Helix ITSM Insights product prior to implementing the product or a feature.
Best practices for proactive problem management
We recommend the following best practices for proactive problem management:
- Select incidents for a specific duration to avoid unnecessary incidents getting clustered.
As issues for each product or service are different, select an appropriate Group by level 1 field based on your organization structure, service ownership, technical stack ownership, or other defining fields.
For better quality clusters with clear boundaries, we recommend adding Group by fields.
Following are some of the common fields used for grouping:
- Service
- Company
- Region
- Product name
- Operational categorization 1, 2 and 3
- Product categorization 1, 2 and 3
- Custom fields
- If you are unable to define cluster boundaries and mixed clusters (across technical stacks, services, or organizations) are acceptable, then you can select only clustering fields. However, this option is not recommended.
When selecting the number of clusters, we recommend selecting the Let the system find no. of clusters check box. The system automatically selects an optimal number of clusters. However, when you know the number of clusters from prior execution runs or domain knowledge, you can specify a value.
- We recommend using stop words for the selected language so that frequently used words in incidents (for example, company name, region) which should not be used for generating cluster names can be ignored.
- Provide appropriate user permissions to users so that they can create problems and analyze incidents.
- Define your schedule based on the problem management process in your organization. The schedule for clustering jobs can be weekly or bi-weekly or monthly based on your review cycles for problem identification. Typically, the lookback period for each of these schedules is the previous 1-4 weeks of data.
- We recommend configuring the one-time job first before configuring the recurring job. If the desired job quality score is achieved, you can replicate the settings in the recurring job.
- After running a job, evaluate the job quality score to determine if the clusters are formed as expected. You can tune the data selection, group by fields, number of clusters, and stop words to achieve the required quality score and the right clusters.
Use filters to select the data you are interested in clustering. We recommend you preview the filters in the Selected filters area before applying them.
- The BMC Helix ITSM Insights algorithm groups incidents based on the fields that you select. For best results in creating clusters, select fields that contain a short description of the incident, and avoid the fields that contain long, unnecessary details.
Best practices for real-time incident correlation
We recommend the following best practices for real-time incident correlation:
- Select appropriate Group by fields for incident correlation configuration based on your company structure, organization structure, service ownership (business service), and ownership of technical stack. Adding group-by fields improves the initial grouping of incidents. We recommend grouping by Assignee - Company (Assigned Support Company), Assignee Support group (Assigned Group) and Company to create clusters with all incidents related to your company and assigned group. Availability of all incidents of your assigned group and company helps you manage their relationships effectively.
- Select textual fields such as Summary, Description as clustering fields.
- To obtain cohesive clusters, we recommend setting the similarity threshold should be in the range of 0.6 to 0.8.
- Avoid changing the configuration often during a run as it clears the existing clusters.
- To improve performance while generating clusters, we recommend excluding the Description (Detailed Description) field from the configuration.
- The BMC Helix ITSM Insights algorithm groups incidents based on the fields that you select. For best results in creating clusters, select fields that contain a short description of the incident, and avoid the fields that contain long, unnecessary details.