What Is Adobe Data Distiller and When to Use It in AEP

You have years of orders, site visits and CRM data in Adobe Experience Platform, but segmentation needs something simpler: how much each customer spent in the last 12 months, which value decile they fall in, or how many people at a B2B account are still active. If that number depends on exporting data to another tool every week, it's slow and brittle. And if you load the full history into the real-time profile instead, Adobe warns that your profile data volume (Total Data Volume) grows without making personalization any better.

At WolfSellers we implement Adobe Experience Platform (AEP), Adobe Real-Time CDP and Adobe Journey Optimizer for companies across North America and Mexico. This article explains Adobe Data Distiller: what it is, how it differs from Query Service, what you can do with it, how it's licensed and when it makes sense. Everything we say about Adobe features comes from Adobe's Experience League documentation, reviewed in October 2026.

---

What is Adobe Data Distiller

Adobe Data Distiller is an add-on package for Adobe Experience Platform that unlocks Query Service batch queries: SQL that cleans, transforms and enriches data that's already been ingested and writes the result back to the data lake as a new dataset, which you can schedule and, if needed, publish to the Real-Time CDP profile. Adobe describes it as a package for post-ingestion data preparation, for real-time profile or analytical use cases.

Three clarifications:

• It isn't a separate tool. It runs inside Query Service, with the same editor and the same SQL. What changes is what you're allowed to do: write datasets, schedule queries and run them for up to 24 hours.
• It doesn't ingest data. It queries what's already in the data lake.
• It isn't a new product. Experience Platform release notes already mention the "Data Distiller SKU" in the October 26, 2022 release. What has changed is what it includes.

Data Distiller vs. Query Service: the difference in one table

Query Service is AEP's SQL engine. It ships with every Experience Platform-based application (Real-Time CDP, Customer Journey Analytics and Journey Optimizer), but only for ad hoc queries. Data Distiller adds batch queries. This is how the Query Service packaging page lays it out:

Permissions differ too: in Adobe Admin Console, Manage Queries without write access lets users run ad hoc queries; with write access, batch queries.

What you can do with Data Distiller

Derived datasets and profile attributes

A derived dataset is a table you build with SQL from other datasets in the data lake: totals, averages, or rankings such as deciles, quartiles and percentiles. You create it with CREATE TABLE ... AS SELECT, append to it with INSERT INTO and schedule it to recompute on its own. To use it in Real-Time CDP it needs a primary identity with its namespace and the profile label, and you can do that in SQL too: WITH (label='PROFILE') when you create the table, or ALTER TABLE ... ADD LABEL 'PROFILE' on an existing one.

Adobe calls this the derive-and-promote pattern: the full history stays in the data lake, which doesn't count toward your profile entitlement, and only the signal (a score, a tier, a rank) moves to the profile. The documented workflow:

1. Write the query that defines the signal, for example 12 months of activity mapped to a loyalty tier.
2. Generate a derived dataset with one row per customer.
3. Schedule a daily or weekly refresh.
4. Publish the dataset to the profile, where the signal shows up in Audience Builder.
5. Activate those audiences through Real-Time CDP destinations or Journey Optimizer journeys.

SQL audiences

Since September 2024, Data Distiller can create audiences directly with CREATE AUDIENCE ... AS SELECT. The audience is registered with a Data Distiller origin, also shows up in Journey Optimizer and activates to file-based destinations such as Amazon S3, SFTP or Azure Blob, with its extra columns available as enriched attributes for filtering. Three rules before you design:

• They expire after 30 days. They're external audiences and Adobe deletes them after that window; if they need to live longer, a scheduled query has to rebuild them.
• They refresh in full with INSERT OVERWRITE INTO, available since April 2025 to Data Distiller customers. INSERT INTO only appends.
• They don't go through the profile and only support top-level attributes, not nested ones.

Statistics and machine learning in SQL

Data Distiller trains, evaluates and applies models without leaving SQL: CREATE MODEL trains, model_evaluate measures performance against a separate dataset and model_predict scores new data. Since April 2025, predictions can be saved with CTAS or INSERT INTO. The documentation covers classification algorithms (logistic regression, decision trees, random forest and Naive Bayes, among others), regression and clustering (K-means and Gaussian mixture models, among others).

If your data science team prefers its own tools, the AI/ML feature pipelines workflow computes features with Data Distiller from a Python notebook, brings them into your machine learning environment and sends the scores back to AEP for Real-Time CDP and Journey Optimizer.

Reporting, dashboards and BI tools

SQL Insights builds custom reporting data models and extends the ones that ship with Real-Time CDP, so you can see profile, audience, journey or consent metrics by your own dimensions, such as business unit. Aggregates load into the accelerated store and show up in AEP dashboards or BI tools; the documentation names Power BI, Tableau, Looker and Apache Superset. That store is for aggregated data only: no raw events, identities or personal data, and it doesn't replace the data lake or a data warehouse.

In the editor, Data Distiller accelerators are parameterized SQL templates maintained by Adobe for common analyses such as funnels, moving averages or audience overlap.

How Data Distiller has changed from 2022 to 2026

Data Distiller has been in Adobe's documentation for at least four years. Here's what the release notes record:

The latest changes are about governance and operations, not a headline feature. In September and October 2026, Adobe updated two guides worth reading before you buy: long-term personalization with Data Distiller, and the one comparing it with Customer Journey Analytics and Federated Audience Composition.

How Data Distiller is licensed and how usage is measured

Data Distiller is an add-on you license from Adobe alongside Real-Time CDP, Customer Journey Analytics or Journey Optimizer. We don't publish pricing: your Adobe contract sets it. What is public is how usage is measured:

Compute hours are a performance guardrail, not a hard stop: if you exceed them, Adobe warns that performance may degrade. You track them in the license usage dashboard (it requires the View License Usage Dashboard permission) and in the scheduled queries tab, which shows compute hours per run from August 15, 2024 onward. What ultimately governs is your sales order and the product description in your contract.

Use cases in retail, ecommerce and B2B

Typical use cases, all following the same pattern: compute in the data lake and send only the result to the profile or the audience.

An illustrative example of the first use case, with made-up dataset and field names:

CREATE TABLE customer_value_12m AS (
  SELECT
    customer_id,
    DATEDIFF(current_date, MAX(order_date)) AS days_since_last_purchase,
    COUNT(DISTINCT order_id) AS orders_12m,
    SUM(order_total) AS spend_12m,
    NTILE(10) OVER (ORDER BY SUM(order_total) DESC) AS spend_decile
  FROM ecommerce_orders
  WHERE order_date >= date_sub(current_date, 365)
  GROUP BY customer_id
)

Decile 1 holds the top 10% of spenders. With a primary identity and the profile label, spend_decile becomes an attribute you can build audiences on.

Two warnings from Adobe. Many datasets, such as those from the Adobe Analytics source connector, already arrive deduplicated at the event level, and doing it again only slows the query down; streamed data does need that step. And with Adobe Target, the Real-Time CDP destination accepts batch and streaming audiences, but what you can personalize depends on your implementation.

When Data Distiller makes sense and when it doesn't

Adobe's guide to choosing a long-term personalization approach starts with one question: does your segmentation look back further than 30 to 90 days? If not, Real-Time CDP and Journey Optimizer are enough. If it does, you have several paths:

Federated Audience Composition requires Real-Time CDP or Journey Optimizer in the Prime or Ultimate package plus its own add-on, and connects to Snowflake, Google BigQuery, Databricks, Amazon Redshift, Azure Synapse, Microsoft Fabric and Vertica. We cover when to ingest and when to federate in data governance in Adobe Experience Platform.

Data Distiller isn't a fit if:

• Nobody on your team writes and maintains SQL. Adobe says so: it requires SQL and a data engineer.
• The analysis is exploratory. The transformation has to be defined up front.
• You need the signal computed at the moment of the event. It runs in batches; that's what streaming segmentation and Journey Optimizer are for.
• You're looking for a data warehouse or a general BI layer.

The tools also work together. Adobe suggests having Data Distiller compute a churn risk score every week and letting analysts in Customer Journey Analytics combine it with the behavior they uncover.

Data governance and privacy in Data Distiller

A derived dataset is still customer data, and the Query Service documentation asks you to govern it like its source:

• Usage labels. Datasets built on a labeled XDM schema inherit its labels. Datasets created by CTAS with an ad hoc schema must be labeled too, which first means making that schema visible in the UI.
• Column-level access. With attribute-based access control, a user without permission on a column can't read or write it. Views (CREATE VIEW) don't inherit those rules.
• Identity for privacy requests. Add an identity to every derived dataset so access and delete requests can reach it. Deleting records from the data lake doesn't delete the profiles built from them.
• Pending deletions. Don't run derived queries on a dataset with a pending deletion: they would capture the data you're deleting.
• Auditing. Audit logs record who ran queries and who created, changed or deleted templates and scheduled queries.

If you operate in Mexico, using personal data to compute scores or segments is a purpose your privacy notice must cover under the LFPDPPP, Mexico's federal data protection law for private companies; we explain how that maps to AEP in data governance in Adobe Experience Platform.

How to get started with Data Distiller: a 4-to-6-week pilot

We recommend starting with a single use case that has a business owner and a clear metric:

The roles: data engineering or a SQL analyst to write and maintain queries; data architecture with XDM experience for schemas, namespaces and identity; marketing to define the signal and use the audience; data governance for labels and access; and AEP administration for permissions and license usage.

Common Data Distiller mistakes

1. Buying it without a defined use case. Compute hours are licensed per year; without a use case, you pay for capacity nobody uses.
2. Loading the full history into the profile instead of the computed signal.
3. Recomputing the entire history on every run. Adobe documents an incremental load pattern with snapshots that processes only what's new.
4. Expecting UPDATE or DELETE. For datasets, Query Service is append-only: you reflect a change by inserting a new record with the correct state.
5. Forgetting to label derived datasets that use ad hoc schemas.
6. Testing heavy queries in development without watching usage, when capacity is shared across sandboxes.
7. Letting SQL audiences expire because nobody scheduled a refresh.
8. Storing personal data in the accelerated store or using it as a data warehouse.

How we do it at WolfSellers

We start with an assessment of your AEP instance: datasets, schemas, identities, ingestion quality and the use cases marketing and sales want to solve. With the business we define the signal and how it will be measured. We write and version the SQL, validate it against reports you already trust and govern it in a development sandbox. Then we schedule it with alerts, estimate compute hour usage, publish it to the profile or as an audience, and document how your team runs it.

If you're still evaluating the platform, start with our Adobe Real-Time CDP guide. We're an Adobe Solution Partner at the Gold level, with a 135-person team working from Mexico City in a nearshore model. Our Adobe Experience Platform, Adobe Real-Time CDP, data engineering and business intelligence services cover design through operations. If you'd like us to look at your case, get in touch.

---

Frequently asked questions about Adobe Data Distiller

Is Data Distiller the same as Query Service?

No. Query Service is Adobe Experience Platform's SQL engine and ships with Real-Time CDP, Customer Journey Analytics and Journey Optimizer, but only for ad hoc queries of up to 10 minutes that don't save results. Data Distiller is the add-on that unlocks batch queries: they write datasets, can be scheduled and run for up to 24 hours.

Do I need Data Distiller to use Real-Time CDP?

No. According to Adobe's guide, if your segmentation looks back 30 to 90 days and the profile has the data you need, Real-Time CDP and Journey Optimizer are enough. Data Distiller pays off when you need signals computed over long histories, such as deciles, scores or tiers.

Is Data Distiller licensed separately?

Yes. It's an add-on you license from Adobe alongside Real-Time CDP, Customer Journey Analytics or Journey Optimizer, measured mainly in annual compute hours. Pricing depends on your Adobe contract.

Is Data Distiller a new product?

No. Experience Platform release notes have mentioned it since October 2022. The newer pieces are SQL audiences (September 2024), full audience refreshes and model predictions saved to tables (April 2025), and session and permission improvements in 2026.

What are Data Distiller compute hours?

They're the package's unit of consumption: the time the Query Service engine takes to read, process and write back to the data lake during a batch query. They're licensed per year, reset on your contract anniversary and show up in the license usage dashboard.

Can I connect Power BI or Tableau?

Yes. Query Service uses the PostgreSQL protocol, and Adobe documents connections for Power BI, Tableau and Looker, among others. With Data Distiller you can prepare aggregated tables in the accelerated store so dashboards respond quickly.

Does Data Distiller replace my data warehouse?

No. It prepares and enriches data that already lives in Adobe Experience Platform. If your history lives in Snowflake, BigQuery or Databricks and must stay there, look at Federated Audience Composition.

Can I use the results in Adobe Journey Optimizer?

Yes. A derived attribute published to the profile works in audiences and journeys like any other, and audiences built with SQL show up in Journey Optimizer, where you can filter them by their enriched attributes.

Related services

If this topic is relevant to your business, these services from WolfSellers can help you implement it:

• Adobe Experience Platform
• eCommerce Marketing
• AI & Automation