Dataset Deletion

27 Sep 2026 » Platform

The last option you have to delete data in the Adobe Experience Platform (AEP) is to delete the data container: the dataset. I know, this sounds very obvious, but there are some wrinkles that you need to be aware of.

If you are not familiar with them, I already have a full post on datasets, which I suggest you review before proceeding with this one. In summary, a dataset is similar to a database table. This is where all the data resides: profile attributes and events. The structure of a dataset is defined by an XDM Schema.

Before You Delete a Dataset

Before you go ahead and delete a dataset, you need to prepare the operation.

System Datasets

I am sure AEP already protects you from deleting a system dataset, although I have never tried to delete one. In general, the guidance is that you should never delete a system dataset. If you must delete one of them, open a ticket with Adobe Client Care.

Data Sources

For obvious reasons, you should stop sending data to a dataset you want to delete. Check your data sources and disconnect any that are currently populating the dataset. I am not sure whether you can delete a dataset that is connected to an active data source, but I am sure there will be issues if you do so and AEP allows it.

Data Usage

If you have the data in AEP, it is very likely that various AEP components and applications are relying on it. You need to put your Sherlock Holmes hat on and track all the uses of the data that is soon going to be missing:

  • Data Distiller. Check your jobs and stop those that touch the dataset, either as a source or a destination.
  • Segmentation. Change or delete any segments that rely on the soon-to-be deleted attributes.
  • Customer Journey Analytics (CJA). If your dataset is being used in any connection, you need to remove it from the connection.
  • Adobe Journey Optimizer (AJO). Perform a thorough inspection of campaigns, journeys and messages, tracking down any reference to the data to be deleted. Update them accordingly.
  • Real-Time Customer Data Platform (RT-CDP). Make sure your destinations do not need the attributes that will no longer be there.

It is worth mentioning that deleting a dataset does not delete the underlying XDM Schema. In fact, it is not possible to delete an XDM Schema that has been enabled for profile. The consequence is that all data attributes defined in the schema will always be available in the AEP UI. For example, you can create a segment based on a value of an attribute that has no data, the UI will not prevent you; however, no profile will match the condition or the condition will be useless.

People

As important as the technology is, people are even more important. This is probably the most difficult part of the process. You can run some scripts targeting the API to check everywhere the dataset or its data is used. However, finding all the internal users who are relying on the data that is going to be deleted is much more difficult.

Once you find them, you need to notify them that you are planning to delete the dataset and explain to them the consequences:

  • Output of Data Distiller jobs will not have some or all the data.
  • Segments based on that data will not be available or will behave differently.
  • Some CJA reports will be missing data.
  • Some AJO campaigns and journeys will behave slightly differently or will have to be stopped completely.
  • Some AJO messages will have to be updated, as the attributes they were inserting into the body will no longer be populated.
  • Some RT-CDP destinations will receive a slightly different audience, or will not receive any new audiences at all.

I am sure you will have to work with them on alternatives to keep the business running. Expect friction and pushback.

Behind the Scenes

You can delete a dataset both through the UI and the API. This is the easy part of this whole process.

Once you submit the request, it will be sent to a queue for processing. Deleting a large dataset will take hours, maybe days, depending on the size. Finally, once the request is processed:

  • All profile fragments coming from the dataset will be completely gone. This also means that older fragments coming from another dataset will now be part of the profiles.
  • Identities that came from the dataset will be deleted from the Unified Identity Service.
  • All links between identities provided by the dataset will be removed from the identity graph.
  • The dataset will no longer be available in the Data Lake.

As you can see, deleting a dataset not only deletes the stored data, but it also impacts other systems.

Batches

I wanted to finalize by briefly mentioning batches. You may be tempted to delete only a batch to achieve the same result as deleting a dataset, but limited to the data that was part of the batch. That is, of course, if you know what data is in each batch.

However, this is not how AEP works. If you delete a batch:

  • The batch records are removed from the Data Lake.
  • The processed data remains in the Unified Profile Store.
  • Identity graph data is not removed by the batch deletion.

 

Photo by Pixabay

Related Posts