Generating metrics from RUM events is in beta. Access to this feature is provisioned to customers using Real User Monitoring. Contact Datadog Support to provide feedback.
Real User Monitoring (RUM) allows you to capture events that occur in your browser and mobile applications using the RUM SDKs and collect data from events at a sample rate. Datadog retains this event data in the RUM Explorer, where you can create search queries and visualizations.
RUM-based metrics are a cost-efficient option to summarize the data from your set of RUM events. With RUM-based metrics, you can visualize trends and anomalies across your RUM data at a granular level for up to 15 months.
You can also generate a count metric of RUM events that match a query or a distribution metric of a numeric value contained in RUM events, such as the request duration.
To create a metric from a search query in the RUM Explorer, click the Export button and select Generate new metric from the dropdown menu.
Select an event type you want to create a metric for, such as Actions. Your options include Actions, Errors, Resources, and Long Tasks. For more information, see Search RUM Events.
Create a search query that filters your RUM events using the RUM Explorer’s search syntax.
Choose a field to track from the dropdown menu next to Count.
Select * to generate a count of all RUM events that match your search query.
Optionally, enter an event attribute such as @action.target to aggregate a numeric value and create corresponding, aggregated metrics such as count, min, max, sum, and avg.
If the RUM attribute facet is a measure, the metric value is the RUM attribute value.
Select a path to group by from the dropdown menu next to group by. The metric tag name is the original attribute or tag name without the @. By default, metrics generated from RUM events do not contain tags unless they are explicitly added. You can use an attribute or tag dimension that exists in your RUM events such as @error.source or env to create metric tags.
RUM-based metrics are considered to be custom metrics. Datadog recommends avoiding grouping by unbounded or extremely high cardinality attributes such as timestamps, user IDs, request IDs, and session IDs. For more information, see Log Management Data Security.
Add percentile aggregations for distribution metrics. You can generate P50, P75, P90, P95, and P99 percentiles.