Snowflake native Experimentation — LaunchDarkly

Contact us for help with configuring Snowflake native Experimentation

If you have both Data Export and Experimentation enabled for your LaunchDarkly account, you should have access to Snowflake native Experimentation. If you do not have access or need help getting started, contact your LaunchDarkly representative or start a Support ticket.

Overview

This category includes documentation about how to connect LaunchDarkly to Snowflake and run LaunchDarkly experiments using metric events from your Snowflake warehouse. You can then export experiment data into Snowflake where your data teams can conduct custom, advanced analysis in addition to the experiment analysis conducted by LaunchDarkly.

Before you can begin running Snowflake native experiments, you must configure the Snowflake Data Export integration in LaunchDarkly and configure the Warehouse Native Experimentation app in Snowflake.

Read the following topics to understand how to configure the required integrations, create a Snowflake native experiment, and analyze the experiment results:

Snowflake native Experimentation architecture

Configuring warehouse native Experimentation sets up three virtual warehouses in your Snowflake instance:

LD_EXPORT_WH
LD_EXPERIMENTATION_WH
LD_SERVICE_WH

Using three warehouses for periodic operations helps reduce compute and gives you visibility into where you are using compute.

The basic data flow process for LaunchDarkly’s warehouse native experimentation is as follows:

End users encounter LaunchDarkly experiments and generate flag evaluations, called “assignment data” in Snowflake. LaunchDarkly SDKs send these events to LaunchDarkly.
Using LaunchDarkly’s Snowflake Data Export integration, the Snowflake LD_EXPORT_WH warehouse pulls flag evaluation and experiment metadata into your Snowflake instance.
The Snowflake LD_EXPERIMENTATION_WH warehouse combines LaunchDarkly flag evaluation and experiment metadata with your Snowflake metric events data to calculate experiment results.
LaunchDarkly uses the Snowflake LD_SERVICE_WH warehouse to periodically sync experiment results from Snowflake to LaunchDarkly. You can view the results in LaunchDarkly on the experiment’s Results tab.

Here is an illustration of the Snowflake native Experimentation architecture:

An illustration of Snowflake native Experimentation architecture.