Architecture guide
The purchase is in AEP. Where does it go missing in CJA?
The team finds a purchase in the dataset but not in Workspace. Repeating the load can duplicate data without solving anything. First find the last place it appears and check how it is counted.
Sources checked:
Follow the purchase into the report
-
AEP dataset
- Purchase identifier
- Received event and fields
Find the purchase using a recognisable key. Check its timestamp, identity and amount. A completed batch does not prove that this row and its fields arrived as expected.
-
CJA connection
- Dataset and import
- History and identity
Check that the dataset is included and the purchase period was imported. Review the selected Person ID and skipped records before blaming the report.
-
Data view and Workspace
- Metric and view rules
- Report dates and filters
Verify the field behind the metric. Compare the time zone, date range and segments on the view, panel and column. The data can be available yet excluded from this reading.
What Adobe documents
Importing new data does not replace historical import
A connection lets you configure new data imports and request historical backfill per dataset. A dataset existing in AEP does not mean its entire history is available in CJA.
Adobe: Create or edit a connectionCheck identity in the connection
The selected Person ID is part of the event dataset configuration. A record without a usable value for that identifier can be omitted from CJA ingestion even when it exists in the AEP dataset.
Adobe: Create or edit a connectionA field and a component are not the same thing
A data view turns fields into dimensions and metrics with their own settings. The same field can feed different components. Validate the component used in the report, not just its displayed name.
Adobe: Create or edit a data viewSegment scope changes what is included
CJA supports person, session and event segments. A purchase condition does not necessarily return the same rows at those levels. Check the scope as well as the condition.
Adobe: Segmentation overviewHow I would review it
- Set the reference: what counts as a purchase, which timestamp and time zone apply, and how cancellations or duplicates are handled.
- Find a known purchase, then compare aggregates with matching scope. Query Service checks the dataset; it does not reproduce CJA settings.
- Separate missing data, skipped data and data excluded from the report. Each cause needs a different correction.
- Record expected differences between platforms. Reprocess only when the cause requires it and after checking for possible duplicates.
This is a diagnostic path, not a guarantee of matching totals across platforms. Identity, sessions and attribution can produce legitimate differences. Do not reload data or retire the previous platform without validating the cause.