Compare data feeds in Customer Journey Analytics and Adobe Analytics
Data feeds in both Customer Journey Analytics and Adobe Analytics allow you to export raw data to third-party platforms.
If you previously used data feeds in Adobe Analytics, use the following information to understand differences in available features and concepts.
For a comparison of data feeds against other Customer Journey Analytics export methods, such as Full Table Export, see Analytics product comparison.
Features available only in Customer Journey Analytics data feeds
The following capabilities are available in Customer Journey Analytics data feeds but are not available in Adobe Analytics data feeds:
-
Derived fields: Custom components built from rule-based transformations that can be included in your data feed schema.
-
Component settings: Data view component settings, such as persistence, metric deduplication, and value bucketing, can transform a component’s value directly in your data feed output, without requiring SQL.
-
Stitching: Cross-device identity resolution that links events across devices to a single person.
-
Structured data model: Feeds are built and delivered using structured data rather than flat strings such as post_product_list. Reflects the existing structure from the XDM schema and the data view.
-
Parquet output: Files are delivered in Parquet format, which natively supports complex nested and structured data. This means that data is easier to access in a database using industry-standard practices.
-
Segmentation: Segments applied to the data view are automatically inherited, and additional segments can be applied directly to the feed.
-
Hive-style partition paths: Output files use Hive-style paths for efficient querying in data lake environments.
-
Component updates apply retroactively: Changes to components in the data view are reflected historically in backfills.
-
Lookups: Classifications are not included with Adobe analytics data feeds. In Customer Journey Analytics, all lookups are embedded directly in the data.
-
Interface that is familiar to Analysis Workspace users: Components come directly from the data view and are also available in Analysis Workspace. You can select dimensions and metrics using the same component rail as Analysis Workspace, rather than a static list of variable names.
-
More persistence models available: There are five different persistence models that can be used in Customer Journey Analytics data feeds.
The Functionality comparison table below covers each of these features in detail, along with differences in features that exist in both products.
Functionality comparison
The following table compares key concepts and configuration options across Customer Journey Analytics data feeds and Adobe Analytics data feeds.
The type of data that can be collected and included in data feeds.
Data is processed at different stages, depending on which product you are using.
Cross-device and cross-channel identity resolution that links events to a single person.
Determines how often the data feed is sent and the window of time included in the feed.
Daily (midnight to midnight in the report suite’s time zone) or Hourly.
15-minute feeds are possible but not available by default.
Late-arriving hits
Hits whose timestamps belong to a previous delivery frequency window but arrive after that window already elapsed.
For example, late-arriving hits could come from a mobile app that buffers events while offline and sends them when it reconnects.
Late-arriving hits can be included or excluded via the Late-arriving hits configuration option.
The Lookback window setting controls how far back the system reaches to include delayed data.
Hits whose timestamps don’t match the order in which they were received.
Because Customer Journey Analytics accepts both streaming and batch data, there is no guarantee that events for a given person will arrive in timestamp order. Although Customer Journey Analytics reorders by timestamp per person, it can only export the data that has arrived. This means that late-arriving hits might be exported after hits with a later timestamp.
The Processing delay setting helps reduce out-of-order events in data feed output by giving more time for batch data to arrive before the export. Event ordering in the delivery is not guaranteed.
Important: The ultimate consumer of your data feed data must be able to handle timestamps that are out of order, per person, because hit ordering in the data feed delivery is not guaranteed.
Exports historical data between two past dates.
The data feed schema determines which columns are available to include in a data feed.
eVar1 / post_eVar1), which accounts for much of the column count.The interface used to configure which columns are included in a data feed.
eVar1, prop5) selected from a pre-defined set of ~1,100+ columns. Components are not named or described beyond their variable identifier.Custom components defined using rule-based transformations applied at report time.
Data view component settings, such as persistence, metric deduplication, and value bucketing, that transform a component’s value at report time.
Whether changes to component configuration are reflected in past and future data feed output.
Lookup datasets in Customer Journey Analytics are the equivalent of classifications in Adobe Analytics.
How a visit or session boundary is defined, which affects how events are grouped and attributed.
The ability to filter data feed output using segments.
Custom metrics that you can create from existing metrics.
How or whether dimension values persist from one event to the next.
The format used for data feed output files delivered to your cloud destination.
Parquet
Natively supports complex nested and structured data. Fields such as post_product_list are represented as structured arrays/nested objects.
Requires a Parquet-aware tool to read, such as BigQuery, Snowflake, or Apache Spark.
The schema structure is embedded within the output file.
TSV
Flat, human-readable rows. Does not natively support structured data; complex fields such as product lists must be encoded as proprietary delimited strings requiring custom parsing logic.
The directory structure used for delivered output files.
year=2024/month=01/day=15/), enabling efficient partition pruning when querying data in data lake environments such as Databricks or Apache Spark.The cloud storage locations where data feed output files can be sent.
Amazon S3, Azure RBAC, Azure SAS, Google Cloud Platform.
Also supports SFTP.
Whether the data feed builder uses the same components and terminology as Analysis Workspace.
The persistence models that are available for dimensions in a data feed.