Amplitude events as a metric source
Register a linked Amplitude project as a Statsig Cloud metric source and build Statsig metrics from Amplitude events.
This feature is in Early Access. During this time, aspects of the functionality may still be developed, and this documentation may not always be up to date. If you have any questions, contact Statsig Support.
A Statsig Cloud project can build metrics from the events in a linked Amplitude project. Statsig translates each metric definition into a query against Amplitude and runs it once a day. The resulting per-user daily values join your experiment exposures through the regular Cloud results pipeline.
Use an Amplitude metric source when your experiments run in Statsig but the events that measure them live in Amplitude. Events move only from Amplitude to Statsig. To send Statsig exposures into Amplitude, use Amplitude Native.
Prerequisites
- You've linked your Statsig organization to your Amplitude organization. Refer to Amplitude for the linking steps.
- The Amplitude project uses the US data center.
- The Statsig project uses Cloud. Warehouse Native projects use a different procedure, described in Warehouse Native projects.
Create the metric source
- In Statsig, go to Data > Metric Sources.
- Click Create.
- Select an Amplitude project. Each Amplitude project supports one metric source, so the picker marks projects that already have one as "Already has a metric source".
- Optionally assign tags and a team.
- Click Confirm.
Statsig configures the rest of the source. The source reads the Amplitude project's event table, uses event_time as its timestamp column, and maps identifiers as follows:
| Amplitude column | Statsig unit ID |
|---|---|
| User ID | User ID |
| Device ID | Stable ID |
To analyze experiments randomized on another unit, add that unit with Add ID Type on the metric source.
Create a metric from the source
- Open the metric source and click Create Metric.
- Enter a metric name and an optional description.
- Click Create.
- Define the metric against an Amplitude event. Choose the aggregation and any event or user property filters.
If your project requires metric reviews, the new definition waits for approval from someone in the metric source's Allowed Reviewers list before Statsig computes the metric.
Supported aggregations
| Aggregation | What Statsig computes per user per day |
|---|---|
| Count | The number of matching events |
| Sum | The sum of a numeric event property |
| Mean | The sum and the count, with the average taken across users |
| Unit Count | Whether the user triggered the event that day |
| Ratio | A numerator and a denominator, as separate columns |
Statsig doesn't support maximum as an aggregation on Amplitude metric sources, because summing daily maximums doesn't produce the maximum for the analysis window.
The numerator and the denominator of a ratio metric must read from the same Amplitude project.
Freshness and time bucketing
Statsig computes metrics from an Amplitude metric source in a daily batch, not in real time. A metric reflects Amplitude data through the most recent completed run. Statsig buckets each day using your organization's fixed UTC offset, so metric days match how Statsig buckets exposures.
Warehouse Native projects
Warehouse Native projects don't query Amplitude directly. Instead, export the Amplitude events into your warehouse and register a standard metric source over the exported table.
- In Amplitude, go to Data and configure an event export to your warehouse. Export every event. To reduce cost, filter the export to only the events you need.
- In Statsig, create a metric source that points to the exported table. Refer to Metric sources for the configuration options.
- Create your metrics from that source in the metric catalog.
This path uses the standard Amplitude event export and the standard Warehouse Native metric source, so it has no additional availability requirements.
Was this helpful?
For AI agents: a documentation index is available at /llms.txt. Append .md to any page URL for markdown, or send Accept: text/markdown.