Setup

Overview

To start using Warehouse Native Experimentation:

1. Connect your data warehouse to Harness FME.

2. Prepare your assignment and metric source tables in your data warehouse.

   • Assignment source tables store raw assignment or exposure events in your data warehouse. They capture how users, accounts, or sessions were allocated to specific experiment variants, along with metadata such as experiment ID, treatment, timestamp, and environment.

     Harness FME reads from these tables to determine which users were exposed to which treatments, ensuring accurate linkage between assignment data and downstream metric analysis.

   • Metric source tables contain raw event-level data used to compute experiment metrics. Each row typically represents a user, session, or account interaction, such as a page view, purchase, or API call—along with associated properties (for example, value, timestamp, or event context).

     Harness FME queries these tables to retrieve and aggregate events for metric definitions, ensuring that experiment analyses are based on consistent, verifiable data directly from your warehouse.

3. Configure your assignment and metric sources in Harness FME.

   • An assignment source defines how Harness FME should read impression/exposure events from your data warehouse and map them to experiments. It ensures that users are correctly assigned to treatments, environments, and traffic types, enabling accurate metric analysis across experiments.
   • A metric source defines how Harness FME reads and interprets raw event data from your warehouse. It ensures that metric events are correctly captured, timestamped, scoped to environments and traffic types, and made available for metric definitions.

4. Define your metrics and create experiments in Harness FME.

Once you've created metric definitions and started running experiments in your data warehouse, you can access analyses in Harness FME.