Instrument Monitor GMAU via aggregate counter in usage ping
Overview
Define and instrument Monitor:Health GMAU which will be based on Incident Management. The proposal and content of this issue was copied from #229918 (closed), which is the strategy of the Plan team for project management.
GMAU definition for Incident management
The sum of all unique users that took an action on an Alert or Incident over a given period of time.
Goal
Track user interactions with alerts and incidents within the usage ping so that we can calculate Monthly Active Users (MAU) for Incident Management which will be the GMAU for Monitor:Health. It is important that we have an accurate reflect of MAU as it is one of the primary metrics that GitLab leadership uses to make informed investment decisions across the product.
Proposal
- Implement a single counter in the usage ping that aggregates the total count of unique users that took an action (listed below) on an alert or incident.
- (might be reprioritized for later) Instrument usage ping counters for all actions individually so that we can understand each action individually.
Record the aggregate unique MAU counts under incident_management_events in the usage ping.
Why are we using a single counter to define GMAU?
For now, we cannot tell unique users across multiple actions if each one is reported in an individual Usage Ping metric. With Usage Ping we are only getting broad signal for "are users interacting", and so the more bundled our counter, the better.
Events To Aggregate In Usage Ping
| Action | Description | Added To Counter |
|---|---|---|
| alert_status_change | Changed status of alert | |
| alert_assigned | Changed assignee on alerts | |
| alert_todo | Added to-do on alert | |
| incident_created | Incident created | |
| incident_reopened | Reopened incident | |
| incident_closed | Closed Incident | |
| incident_assigned | Changed assignee on incidents | |
| incident_todo | Added to-do on Incidents | |
| incident_comment | Comment added to incident | |
| incident_zoom_meeting | Zoom meeting associated with incident | |
| incident_published | Incident published to status page | |
| incident_relate | Marked Incident as related to another | |
| incident_unrelate | Removed relation on incident | |
| incident_change_confidential | Marked the incident as confidential or non-confidential | |
Identifiers
Totals are unique users over all the events tracked in the same category, in this case incident_management. Totals metrics have format #{category}_total_unique_counts_weekly and #{category}_total_unique_counts_monthly
For Incident management, the identifiers are as follows:
- Weekly: incident_management_total_unique_counts_weekly
- Monthly: incident_management_total_unique_counts_monthly