Profiles getting reduced while using Audiences in AEP

[ShivaniM]

Hi Team,

We created a temporary table in snowflake with 2k records ingested and one audience with same number of qualified profiles.

Then we imported the profiles using source dataflow, and created a segment including the same segment we created in datahub (using snowflake audience's folder).

In dataset we see all 2k records are ingested, however in audience only half profiles are qualifying, 1k.

What could be the possible reasons.

Thanks!

[hemavars]

Hi @ShivaniM , two things -

1. What identity did you onboard the data with? Can you check if your profiles are getting stitched based on more than one identity?

2. Can you share a screenshot of the segment definition?

[ShivaniM]

Hello, 

Upon checking with Adobe team I found unlike 'ups' segments of AEP, segments from datahub rewrites the data if new batch enters AEP. and hence those profiles were no longer part of that audience.

To ensure all profiles are part of that segment, we have added and upsert function in dataset, now the new instance and older instances all stay and qualified profile count is also same.

Thank you.!

[Prateek-Garg]

Is there any documentation to add this "upsert function" in dataset?

[_Manoj_Kumar_]

Hello @Prateek-Garg 

You can refer to the documentation links given below:

Enable a Dataset for Profile Updates using APIs | Adobe Experience Platform

[Prateek-Garg]

Hi Manoj,

Thanks for the documentation link. I went through it and not able to understand the use-case for setting "isUpsert:true" flag for dataset. In what scenario should we set this flag.

Also, this flag is not available from UI to make a change for the dataset.

[abhinavbalooni · Accepted solution]

Hi @Prateek-Garg 

You would set the flag to true where you need to make partial updates to an existing profile or else you would end up overwriting the existing information with what you put in through the latest information.

For instance, for a profile, if you bring in data from CRM from different tables using batch, like loyalty data, preferences data etc, you would end up losing information from the earlier batch if you do not have upsert enabled for a dataset.

Hope the above helps.

Cheers,

Abhinav

[Prateek-Garg]

Thanks Abhinav.

[Prateek-Garg]

Hi @abhinavbalooni , 

Is this "isUpsert:true" flag helps making partial updates to only to profiles Or also at data lake level?

Regards,

Prateek

[abhinavbalooni]

Hey @Prateek-Garg 

In AEP data in data lake sits in form of different datasets as per my understanding and also as documented here https://experienceleague.adobe.com/docs/experience-platform/catalog/datasets/overview.html?lang=en

Since the above flags need the dataset to be upsert enabled so this should reflect in the data lake as well.