Configuring Assignment Sources for Warehouse Native Experimentation
Learn how to configure your assignment source tables in your data warehouse for Warehouse Native Experimentation.
An Assignment Source defines where Harness FME reads experiment exposure (or impression) data from your data warehouse.
When creating or editing an Assignment Source, navigate to the Experiments page from the FME navigation menu and click the Assignment Sources tab. Then, click + Create assignment source.
Name: Enter a name for the assignment source.
Owners: Assign one or more owners to make clear who is responsible for maintaining the assignment source.
Description: Enter a description for the assignment source.
If you are using feature flags in Harness FME, you can export impression data from Harness FME to your warehouse and use that table as the assignment source for Warehouse Native Experimentation. See Sending FME impressions to your warehouse for supported integrations and setup instructions.
To define the assignment source table, select Table name or SQL query in the Source table section.
Select a table
Select an existing table name directly from the schema.
Click Test connection to validate that Harness can query the table successfully before continuing.
With Assignment Sources configured, you can confidently create experiments, knowing all exposures are correctly captured, standardized, and reusable across analyses.
Use a custom SQL query
You must have permissions to access all tables referenced in your query, based on the role and credentials configured when setting up your warehouse connection.
Write a SQL query that outputs the required fields.
After entering your query, click Run query to validate and preview results before proceeding.
Harness shows a preview of the data returned from your table or query so you can validate that the expected rows and columns are present.
Add field mappings
Define the following fields from your assignment source to Harness FME:
Unique Identifier
Maps to the column representing the unique key for user, account, or entity.
Impression Timestamp
Maps to the column representing when the user was assigned to a treatment.
Treatment
Maps to the column that stores the treatment or experiment variant (e.g., control, variant_a).
Configure your environments
Select an environment column and map its values to Harness FME environments. For example, select the ENV_NAME column and map its values (US-Prod, UK-Prod) to your Harness project’s Production environment and map the Stg values (US-Stg, UK-Stg) to your Harness project’s Staging environment.
This allows a single Assignment Source to span multiple environments.
Instead of selecting a column, set a fixed Harness FME environment for the entire Assignment Source (e.g., always Production).
This is recommended if the entire source table is scoped to one environment.
Configure your traffic types
Similar to environments, traffic types can be set up in two ways:
Select a traffic type column (e.g., ttid) and map its values to Harness FME traffic types (e.g., user, account, or anonymous).
This is recommended if the same Assignment Source covers multiple population types.
Instead of selecting a column, set a fixed Harness FME traffic type for the entire Assignment Source (e.g., always account).
This is recommended if the entire source table is scoped to one population type.
Add custom fields
Click + Add new custom field to map any additional columns from your assignment source that are not included in the required field mappings. Select a column from the dropdown menu and enter a label for the field. Then, click Save.
Custom fields can be used to filter data when creating or analyzing experiments, for example, using the experiment_id or targeting_rule.
Manage assignment sources
Assignment Sources can be reusable and standardized, or tailored to individual experiments depending on your organization’s needs:
Reusable, standardized sources are recommended if you have a general impressions/exposures table.
This approach makes setup faster and consistent across teams. Be mindful of potential query processing speed and warehouse costs when working with very large shared tables.
Custom per-experiment sources are recommended if you want to scope data more tightly for privacy, relevancy, or performance.
Limits experiment creators to a specific subset of data, reducing query volume and potential data access concerns.
Ultimately, it’s up to your organization whether to centralize around a single reusable source or create smaller, experiment-specific sources. Many teams use a mix of both strategies depending on scale and governance needs.
Once you've set up the assignment sources that best fit your workflow, you can manage them directly in Harness FME.
Edit: You can update the table reference, query, or mappings as your data model evolves. Changes to an existing Assignment Source may disrupt any experiments that are actively using it.
Delete: Remove outdated or misconfigured sources to reduce clutter and prevent accidental use.
DELETING AN ASSIGNMENT SOURCE
Deleting an Assignment Source that is currently used by an experiment will stop all calculations for that experiment. Historical results remain available, but the experiment will no longer update with new data.
If you want to continue the experiment, recreate the Assignment Source with the same configuration (pointing to the same table in your warehouse), then create a new experiment using that source. Assignment Sources cannot be replaced for an existing experiment.
Troubleshooting
If you encounter issues when configuring an Assignment Source:
Last updated
Was this helpful?