Zero-copy personalization with your data warehouse

Keep audience logic in your data warehouse or CDP, and use Bloomreach to personalize customer communication across channels. You move only the data that each use case needs.

Choose the pattern that fits each use case. Send audiences you build in your warehouse or CDP as external segments, and ingest only the attributes, scores, events, and catalogs you need for activation.

This page explains how data reaches Bloomreach, what zero-copy personalization means, and which warehouses you can connect.

Two ways to work with your data warehouse

You can use your warehouse with Bloomreach in 2 ways. Both are valid, and most customers use both, choosing per use case.

• Compute in your source: Your data team builds models, audiences, and scores in your source, such as a data warehouse or CDP, and sends only the results to Bloomreach. Bloomreach caches the results for activation. Events can expire, while customer properties and external segments are stored. This works best when your data team owns the models. Examples include customer lifetime value (CLV) tiers, propensity scores, and loyalty status.
• Compute in Bloomreach: You send raw events, transactions, and catalog data, and Bloomreach builds segments, predictions, and real-time triggers. This works best when marketers want to self-serve. Examples include post-purchase journeys, browse abandonment, and live app behavior.

Move only the data that each use case needs.

What zero-copy personalization means

With zero-copy personalization, you keep the source of truth in your warehouse or CDP. Bloomreach reads or receives only the data each use case needs and uses it to personalize communication. The type of data you bring in decides whether it expires or is stored.

Bloomreach caches the data it needs, so it can combine that data with consent, frequency settings, and live customer behavior in the moment.

How it works

1. Query your warehouse. Bloomreach reads the tables, views, or queries you choose.
2. Cache only what you need. Bloomreach stores the ingested data that the use case requires.
3. Combine in the moment. Bloomreach combines the cached data with consent, frequency settings, and live behavior.
4. Activate, then expire if you want. You use the data in campaigns. Events can expire when you choose.

Audiences you build in your warehouse or CDP work differently. Your source sends the segment membership to Bloomreach and keeps it updated with changes.

Choose how to store your data

The type of data you bring in decides how long it stays in Bloomreach:

• Events: Can expire after a period you set for each event type, so they suit short-lived data. Learn how to set data expiration.
• Customer profiles: Stored.
• Catalogs: Stored.
• External segments: Stored as customer properties. They don’t expire, so delete the matching property to remove one.

To keep data short-lived, ingest it as events and set an expiration period for that event type.

Example: this weekend's top fans

Your data team computes team affinity in the warehouse. You ingest it daily as an event and set a 3-day expiration for that event type. You use it in Friday’s email and SMS campaigns, and each event expires on its own after 3 days.

Send audiences you build in your warehouse or CDP

If your data team defines audiences in the warehouse or CDP, you can bring them into Bloomreach as external segments, without ingesting the underlying data. Your segment logic stays in your source, and you use the audience in campaigns and scenarios like any segment you build in Bloomreach.

External segments stay up to date. You send the audience once, then send only membership changes as customers join or leave it. You can send audiences from any system that can call the external segments API, not only from Snowflake or Databricks.

When to use zero-copy personalization

Use it when you want your warehouse or CDP to stay the source of truth, such as for short-lived data or audiences you compute elsewhere. You choose which data Bloomreach receives, so you can limit it to what each use case needs. Events can expire after a period you set, while customer profiles, catalogs, and external segments are stored. For other use cases, you can send raw data and let Bloomreach compute segments and predictions. Most customers use both.

Warehouses you can connect

Bloomreach connects to your warehouse in the way that fits it. Snowflake, Databricks Lakehouse, Amazon Redshift, and BigQuery connect directly. Snowflake and Databricks ingest customers, events, and catalogs as often as every 15 minutes. Microsoft Fabric connects through files in Azure Storage and SFTP. Databricks CustomerLake sends audiences as external segments.

Before you start

These points apply when Bloomreach ingests data from your warehouse. They don’t apply to external segments.

• Ingestion runs on a schedule. Data arrives as often as every 15 minutes, not in real time.
• Ingestion uses your compute. Each ingestion runs on your own warehouse, so your schedule and data scope control your costs.
• Health checks can add cost. Bloomreach checks the connection about every 3 hours. If your warehouse auto-suspends, a check can wake it.
• Allow Bloomreach IP addresses. If you restrict access by IP, add your outbound IP addresses to your warehouse's allowlist. See Outbound IP addresses.
• Deletes don't sync. Sync doesn't carry over records you delete.

Send engagement data back to your warehouse

Scheduled Data exports send engagement data to file storage, so your data team can load it into your warehouse for analysis and model training. You can export to SFTP, Google Cloud Storage, Amazon S3, or Azure Storage. Bloomreach can’t export directly to a database.

For Snowflake, you export to Google Cloud Storage and then load the files into Snowflake. See Export data to Snowflake.