Create merge policies

In this lesson, you will create merge policies to prioritize how multiple data sources merge into profiles.

Adobe Experience Platform enables you to bring data together from multiple sources and combine it to see a complete view of each individual customer. When bringing this data together, merge policies determine how data is prioritized and what data is combined to create that unified view.

We’ll stick to the user interface for this lesson, but API options also exist for creating merge policies.

Data Architects will need to create merge policies outside of this tutorial.

Before you begin the exercises, watch this short video to learn more about merge policies:

Permissions required

In the Configure Permissions lesson, you set up all the access controls required to complete this lesson.

About Merge policies and Union Schema

You may recall, in the lesson on batch ingestion, we uploaded two records with slightly different information for the same customer. In the Loyalty data, the customer’s first name was Daniel and he lived in New York City, but in the CRM data the customer’s first name was Danny and he lived in Portland. Customer data changes over time. Perhaps he moved from Portland to New York City. Other things change too, such as phone numbers and email addresses. Merge policies help you decide how to handle these types of conflicts when two data sources give different information for the same user.

So, why did Danny win out as the first name? Let’s take a look:

  1. In the Platform user interface, select Profiles in the left navigation
  2. Go to the Merge Polices tab
  3. The default Merge Policy is timestamp ordered. Because you uploaded the CRM data after the Loyalty data, Danny won out as the first name in the profile:

Merge Policy screen

When multiple schemas are enabled for profile, a Union Schema is automatically created for all profile-enabled, record schema sharing a base class. You can view the Union Schemas by going to the Union Schema tab.

Merge Policy screen

Note that there isn’t a union schema for the ExperienceEvent class. While ExperienceEvent data still lands in profile, because it is time-series based, each event includes a timestamp and id and collisions are not a problem.

Now what if you don’t like that default merge policy? What if Luma decides their CRM system should be the source of truth when there is a conflict? For that, we will create a merge policy.

Create a merge policy in the UI

  1. On the Merge Policies screen, select the Create Merge Policy button on the upper-right
  2. As the Name, enter Loyalty Prioritized
  3. As the Schema, select XDM Profile (note that your custom class—since it is record data—is available for merge policies, too)
  4. For Id Stitching, select Private Graph
  5. For Attribute Merge, select Dataset precedence
  6. Drag-and-drop Luma Loyalty Dataset and Luma CRM Dataset to the Dataset panel.
  7. Make sure Luma Loyalty Dataset is on top by drag and dropping it above the Luma CRM Dataset
  8. Select the Save button

Merge Policy

Validate the merge policy

Let’s see if the merge policy is doing what we would expect:

  1. Go to the Browse tab
  2. Change the Merge policy to your new Loyalty Prioritized policy
  3. As the Identity namespace, use your Luma CRM Id
  4. As the Identity value use 112ca06ed53d3db37e4cea49cc45b71e
  5. Select the Show profile button
  6. Daniel is back!

Viewing a profile with a different merge policy

Create a merge policy with limited datasets

When creating Merge policies using dataset precedence, only the datasets of the same base class that you include in the right are included in the profile. Let’s set up another merge policy

  1. On the Merge Policies screen, select the Create Merge Policy button on the upper-right
  2. As the Name, enter Loyalty Only
  3. As the Schema, select XDM Profile
  4. For Id Stitching, select None
  5. For Attribute Merge, select Dataset precedence
  6. Drag-and-drop only the Luma Loyalty Dataset to Selected Dataset panel.
  7. Select the Save button

Loyalty Only Merge Policy

Validate the merge policy

Now let’s look at what this merge policy does:

  1. Go to the Browse tab
  2. Change the Merge policy to your new Loyalty Only policy
  3. As the Identity namespace, use your Luma CRM Id
  4. As the Identity value use 112ca06ed53d3db37e4cea49cc45b71e
  5. Select the Show profile button
  6. Confirm that no profiles are found:
    Loyalty Only no CRM Id lookup.

CRM Id is an identity field in the Luma Loyalty Dataset, but only primary identities can be used to look up profiles. So, let’s look up the profile using the primary identity, Luma Loyalty Id"

  1. Change the Identity Namespace to Luma Loyalty Id
  2. As the Identity value use 5625458
  3. Select the Show profile button
  4. Select the profile id to open the profile
  5. Go to the Attributes tab
  6. Note that other profile details from the CRM dataset, such as the mobile phone number and email address are not available because we only
    CRM data is not viewable in the Loyalty Only policy
  7. Go to the Events tab
  8. ExperienceEvent data is available despite not explicitly including it in the merge policy datasets:
    Events are viewable in the Loyalty Only policy

More about merge policies

In the profile search, change the merge policy used back to Default Timebased and select the Show profile button. Danny is back!

Viewing a profile with a different merge policy

What is going on here? Well, profile merging is not a one time thing. Real-time customer profiles are assembled on the fly, based on various factors, including which merge policy is used. You can create multiple merge policies to use in different contexts, depending on which view of the customer you want.

A key use case for merge policies is for data governance. For example, say you ingest third-party data into Platform which cannot be used for personalization use cases, but can be used for advertising use cases. You can create a merge policy that excludes this third-party dataset and use this merge policy to build segments for your advertising use cases.

Additional Resources

Now let’s move on to the data governance framework.

On this page