Combined event datasets
When you create a connection, Customer Journey Analytics combines all schemas and datasets into a single dataset. This ‘combined event dataset’ is what Customer Journey Analytics uses for reporting. When you include multiple schemas or datasets in a connection:
- Schemas are combined. Duplicate schema fields are merged.
- The ‘Person ID’ column of each dataset are merged into a single column, regardless of their name. This column is the foundation of identifying unique persons in Customer Journey Analytics.
- Rows are processed based on timestamp.
- Events are resolved down to the millisecond level.
Example
Consider the following example. You have two event datasets, each with different fields containing different data.
example_id
timestamp
string_color
string_animal
metric_a
user_310
1 Jan 7:02 AM
Red
Fox
user_310
1 Jan 7:04 AM
2
user_310
1 Jan 7:08 AM
Blue
3
user_847
2 Jan 12:31 PM
Turtle
4
user_847
2 Jan 12:44 PM
2
different_id
timestamp
string_color
string_shape
metric_b
user_847
2 Jan 12:26 PM
Yellow
Circle
8.5
user_847
2 Jan 1:01 PM
Red
alternateid_656
2 Jan 8:58 PM
Red
Square
4.2
alternateid_656
2 Jan 9:03 PM
Triangle
3.1
When you create a connection using these two event datasets, the following table is used for reporting.
id
timestamp
string_color
string_animal
string_shape
metric_a
metric_b
user_310
1 Jan 7:02 AM
Red
Fox
user_310
1 Jan 7:04 AM
2
user_310
1 Jan 7:08 AM
Blue
3
user_847
2 Jan 12:26 PM
Yellow
Circle
8.5
user_847
2 Jan 12:31 PM
Turtle
4
user_847
2 Jan 12:44 PM
2
user_847
2 Jan 1:01 PM
Red
alternateid_656
2 Jan 8:58 PM
Red
Square
4.2
alternateid_656
2 Jan 9:03 PM
Triangle
3.1
This combined event dataset is what is used in reporting. It does not matter which dataset a row comes from; Customer Journey Analytics treats all data as if it is in the same dataset. If a matching Person ID appears in both datasets, they are considered the same unique person. If a matching Person ID appears in both datasets with a timestamp within 30 minutes, they are considered part of the same session.
This concept also applies to attribution. It does not matter which dataset a row comes from; attribution works exactly as if all events came from a single dataset. Using the above tables as an example:
If your connection only included the first table and not the second, pulling a report using the string_color
dimension and metric_a
metric using last touch attribution would show:
However, if you included both tables in your connection, attribution changes since user_847
is in both datasets. A row from the second dataset attributes metric_a
to ‘Yellow’ where they were previously unspecified:
Cross-channel analysis
The next level of combining datasets is cross-channel analysis, where datasets from different channels are combined, based on a common identifier (person ID). Cross-channel analysis can benefit from stitching functionality, allowing you to rekey a dataset’s person ID so the dataset is properly updated to enable a seamless combination of multiple datasets. Stitching looks at user data from both authenticated and unauthenticated sessions to generate a stitched ID.
Cross-channel analysis allows you to answer questions such as:
- How many people begin their experience in one channel, then finish it in another?
- How many people interact with my brand? How many and what types of devices do they use? How do they overlap?
- How often do people begin a task on a mobile device and then later move to a desktop PC to complete the task? Do campaign click-throughs that land on one device, lead to conversion somewhere else?
- How does my understanding of campaign effectiveness change if I consider cross-device journeys? How does my funnel analysis change?
- What are the most common paths users take from one device to another? Where do they drop out? Where do they succeed?
- How does the behavior of users with multiple devices differ from the users with a single device?
For a more information on cross-channel analysis, refer to the specific use case:
For a more in-depth discussion stitching functionality, go to: