Campaings and Users Raw data Export

Created: 21.02.2023

Updated: 13.10.2023

Author: Alexander Sugakov

Data Retrieval

Data is exported uon request for a specific period to a folder with web authentication access. Separate files are created for each day within the folder. The export is done in CSV format (comma-separated).

One-time exports are stored in a subfolder /raw_data/.

Additionally, all clients have the option to enable automatic daily exports. Daily exports are stored in a subfolder /daily/. The data retention period for daily exports is 7 days, meaning that each day the data file from 8 days ago is deleted.

General Description

The exported files contain a stream of data obtained from integrations. All data is structured in the form of a flat table, where various types of events are listed, each of events has its own set of properties.

There are 7 types of events in the data, identified by the numbers in the 'type' column:

0 - Page View,

1 - Event Hit - an event sent to the platform,

2 - Variation Impression - impression of a variation,

3 - Variation Click - click on a variation (only for campaign types that track clicks),

4 - Widget Impression - impression of a recommendation widget,

5 - Product Impression - impression of a product in a recommendation widget,

6 - Product Click - click on a product in a recommendation widget.

7 - Visible Impression - campaign appearing in the user's viewport

Data Processing Recommendations

Since all user interactions with customers’ entities (website, application, etc.) are stored as separate rows, to obtain a complete list of events for a client, it is necessary to filter events based on the 'uid' column. To obtain a full list of a client's events within one session, you additionally need filter by the 'sessionId' column.

Therefore, to obtain customer purchases attributed to a campaign, it is necessary to select all unique customer’s purchases (to avoid duplicates) made after the Variation Impression event with the corresponding 'variationId' within the same session as the Variation Impression. You can use columns with campaigns, experiences, versions, and variations IDs and names to ideintify specific versions of specific campaigns.

It is recommended to analyze campaigns only within the scope of versions/tests. When changing experience settings or the distribution of users among variations (without adding/removing variations), all campaign IDs, experiences, and variations will remain unchanged. Only the version/test ID will change, but the distribution of variations among users will start anew, and the test concept itself may undergo significant changes.

We aslo typically recommend excluding outliers from calculations.