Audience Count Suddenly Drop (Data Quality Issue)

Overview

This article explains the cause and resolution for workflows that succeed but result in a large, unexpected drop in the final audience segment count (e.g., from 1.5 million customers to 200,000 customers).

This typically indicates a problem with the source data quality or upstream synchronization, not a failure in Amperity's segmentation logic.

1. Symptoms

You observe a sudden and large drop in the customer count (often 50% or more) for a key segment (e.g., "Active Email Subscribers").

  • Workflow Status: The campaign or workflow runs successfully (status "Completed"), and there are no critical Amperity system errors.
  • Segment Behavior: The segment accurately filters the audience, but the result is much lower than expected.

2. Root Cause: Missing Critical Data (External Issue)

Amperity segments only include customers that meet all criteria. The drop occurs because a required data point (like an opt-in status or transaction flag) has gone missing or been corrupted in the source tables that feed the segment.

Common Scenarios:

The data drop is often linked to synchronization failures from external systems, such as:

  • Marketing Cloud Opt-In Status: The segment relies on a field like email_opt_in_status. If the upstream system (e.g., Salesforce Marketing Cloud or Klaviyo) fails to send this data via the Snowflake share, Amperity assumes the status is missing or false, and filters the profiles out.
  • Source Data Corruption: A recent upstream change in a job or view caused a large volume of records in a key table to lose a required identifier (like an amperity_id or email).

How Amperity Reflects the Issue

The Amperity platform works correctly by accurately reflecting the missing data. If your source data table shows only 200K valid records, the segment will only output 200K valid customers.

3. Resolution Steps (Action Required by Your Data Team)

To fix this issue, your internal data team or administrator must locate and repair the upstream data flow.

Step 1: Locate the Data Drop

  1. Identify the Source Table: Note the main table feeding the segment (e.g., Customer_360_SFMC_EMEA).
  2. Review Database Run History: Navigate to Databases and view the run history for that source table.
  3. Pinpoint the Date: Look for the specific date where the Record Count in the run history suddenly dropped. This date marks when the bad data entered the Amperity system.

Step 2: Repair the External Source

  1. Data Source Investigation: The team responsible for your data pipeline (Snowflake/Data Engineering) must investigate the external source system (e.g., the S3 bucket, the SFMC API sync, or the Snowflake view) for the date identified in Step 1.
  2. Restore/Correct Data: The source records must be corrected and then re-sent to the Amperity tenant via the standard ingestion process. This ensures the full, valid dataset is restored.

Once the upstream data is successfully ingested, the segment will repopulate to its original count on the next scheduled workflow run.

If you have any trouble diagnosing the drop point in Step 1, please contact Amperity Support, and we can provide the specific table run history details.