Integrating with Braze

Set up A/B test in Braze and analyze in Eppo

Braze’s built in A/B testing for campaigns combined with Eppo’s data warehouse native integrations allow for teams to measure downstream metrics such as revenue or retention. This combination of tools gives teams richer insights into their CRM and Lifecycle Marketing campaigns.

In this section, we will walk through how to set up your analysis in Eppo if you use Braze for experimentation. This can be especially helpful to users who want to leverage Braze’s UI to easily setup a test without additional code and compare test results using data warehouse metrics that cannot be shared with Braze. More specifically, this guide will provide instructions on how to:

• Integrate your Braze data to your warehouse
• Set up your experiment in a Braze Campaign or Canvas
• Set up Eppo to measure Braze experiments

Braze Setup

Integrate Braze campaign data to warehouse.

Since Eppo can analyze experiments directly in your warehouse, all that is needed is the Braze campaign data in the warehouse that you have connected to Eppo. Braze allows you to export campaign data using Currents. Additionally, Braze data can be made available in your Snowflake instance through Braze Data Sharing.

Creating experiments in Braze

Braze has two main paths for creating an experiment that will have different downstream data schemas associated with them:

1. Canvas -- Multi-step campaign builder that allows users to test different steps throughout a user's journey. See Setting up a Variant in a Canvas in Braze's documentation.
2. Campaign -- Send one off messages through a variety of channels available through Braze. See Creating multivariate and A/B tests in a Campaign

Create a Canvas in Braze

Follow Braze's documenation for setting up an experiment within Braze. This will include:

• Adding a variant within your Canvas.
• Determining the steps that should be included in each variant.
• Who is eligible for this Canvas.
• Scheduling when this Canvas should run.

Create a Campaign in Braze

Follow Braze’s documentation for setting up your experiment within a Campaign. This will include:

• Selecting the message type (e.g., Push, Email, In-App Message, etc.).
• Defining the target audience for your A/B test.
• Configuring the variations.
• Scheduling the campaign.

Once your experiment is set up, launch the test in Braze.

Eppo Experiment setup

The first step to setting up your experiment in Braze is to create an Assignment table in Eppo based on user-level Braze campaign data in your warehouse. We recommend setting up separate Assignments table for Canvas experiments and Campaign experiments since each will use different metadata in the Braze event payload.

Use Braze’s docs to determine which event schema you should use as experiment assignments in Eppo. More context about the specific event types to use are provided below.

Canvas Assignment Table Setup

For Canvas experiments, there are two ways to create experiments:

1. At the Canvas level that corresponds to the users.canvas.Entry event
2. Within the Canvas that corresponds to the users.canvas.experimentstep.SplitEntry event

For experiments on the Canvas entry, canvas_name is used as the Experiment name, and the canvas_variation_name as the Variation name in the Assignments table.

For experiments witin a Canvas, experiment_step_id is uesed as the Experiemnt name and experiment_split_id is used as the Variation name in the Assignment table.

Additionally time as the Timestamp for both of these experiment types on the Eppo Assignments table.

These events in Braze represents when a user enters an Experiment step or a Canvas and the timestamp will be the most accurate representation of when a user was exposed to the experiment.

Your Braze Canvas Assignments table will look like this:

Column	Type in Eppo	Data from Braze Table
assignment_timestamp	Timestamp	time
user_id	Experiment subject ID	user_id This should be an ID that maps to the metrics used in a experiment, this will most likely be an internal User ID
experiment	Experiment key	canvas_name and experiment_step_id
variant	Variant	canvas_variation_name and experiment_split_id

See below for a visual overview of what Braze Canvas steps map to as Braze data.

Campaign Assignment Table Setup

For Campaign experiments, you will identify the channels your team uses. Typically we recommend using Send events to represent when a user was entered into an experiment. The following Send events all have a campaign_name to represent the Experiment name, message_variation_name to represent the Variation name, and time to represent the timestamp in an Eppo Assignment table:

• users.messages.contentcard.Send
• users.messages.email.Send
• users.messages.pushnotification.Send
• users.messages.sms.CarrierSend
• users.messages.sms.Send
• users.messages.whatsapp.Send

Your Braze Campaign Assignments table will look like this:

Column	Type in Eppo	Data from Braze Table
assignment_timestamp	Timestamp	time
user_id	Experiment subject ID	user_id This should be an ID that maps to the metrics used in a experiment, this will most likely be an internal User ID
experiment	Experiment key	campaign_name
variant	Variant	message_variation_name

Including Campaign and Canvas Specific Metrics with a Secondary Entity

When analyzing channel outreach experiments, it's common to measure both overall user-level metrics (engagement, revenue, retention, etc.) and campaign-specific metrics (click through rate, unsubscribes, etc.).

A Secondary Entity on the Assignments table will ensure any message-specific metric such as Clicks, Bounces, or Opens are specific to the Campaign or Canvas experiment.

To include Campaign or Canvas specific metrics:

1. Create a new Entity called Combined ID.
2. On the Braze Campaign or Canvas Assignments table, concatenate the User ID and Campaign or Canvas Name and adding it as a Secondary Entity called combined_id For example, a Campaign called Email_14_day_nurture, would be concatenated to 12345-Email_14_day_nurture as the combined_id.
3. Include this combined_id on the Assignments table as a Secondary Entity. Your Assignments table should look like this:

Column	Type in Eppo	Data from Braze Table
assignment_timestamp	Timestamp	time
user_id	Experiment subject ID	user_id Internally defined User ID
experiment	Experiment key	campaign_name, canvas_name, experiment_step_id
variant	Variant	message_variation_id, canvas_variation_name, and experiment_split_id
combined_id	Combined ID Entity as the Secondary ID	Concatentated campaign_name, canvas_name, or experiment_step_id with user_id

4. Concatenate the Campaign or Canvas Name with the User ID on your Braze Facts table and use that as the Entity for the Fact. Your Fact table should look like this:

Column	Type in Eppo	Data from Braze Table
timestamp	Timestamp	time
user_id	No column specified	user_id Internally defined User ID
experiment	No column specified	campaign_name, canvas_name, experiment_step_id
combined_id	Combined ID as the Entity ID	Concatentated campaign_name, canvas_name, or experiment_step_id with user_id

4. Create Metrics from your newly created Braze Fact tables.
5. When setting up an experiment for analysis, make sure you select the Combined ID as the Secondary ID used for Analysis. Consider setting up a Protocol for Braze experiments to automate this experiment setup for subsequent experiments.