Imports Overview (Deprecated)

Overview of data import options for Statsig, including ingestion from data warehouses, object storage, and customer data platforms via connectors.

This solution is still functional, but can be manual and time consuming to set up with minimal error handling. Check out the Data Warehouse Ingestion solution instead.

How imports work

Statsig supports several ways to send your own events or pre-computed metrics. Statsig can read directly from your data warehouse through native integrations, or you can write your data to Azure storage owned by Statsig.

How metric imports work

You write your metrics data at a user-day-metric granularity to a table (or API endpoint) with a specific schema. You write dates as rows to a signal table (or API endpoint) after you finish adding metrics for that day.

Statsig polls this signal table hourly. After you mark your first day of data as ready, or mark the next day of your data as ready, Statsig pulls data from your table and processes it into metrics and experiment results.

Notes:

New imported metrics may not appear in your catalog until the following morning.
The earliest signal date Statsig considers for your first day of imported metrics is 4 weeks ago.
Statsig doesn't show metrics that haven't had a value for 2 weeks in the Metrics Catalog. Start your imports on a recent date for them to appear.
Statsig imports dates in order without gaps. If you update the signal table for 2022-06-13 and 2022-06-15, Statsig doesn't load data for 2022-06-15 until you mark 2022-06-14 and Statsig finishes ingesting that data.

How event imports work

Event imports work similarly to metric imports, with a schematized data source you own and triggers for ingestion. Unlike metric imports, Statsig pulls event data continuously to avoid large batch loads timing out.

Support

Contact Statsig in Slack if you have issues with imports or feature requests.

Was this helpful?