Action Exports Overview

Purpose

This article provides a breakdown of action export and response export data, focusing on the distinction between session-level and action-level reporting. Understanding this is important for calculating revenue and reconciling with InnovidXP. It also explains the differences between data in action exports and in the platform.

Overview

Action exports provide action-level data from InnovidXP, so this level of reporting differs from other views in the platform, such as in the Pivots section.

For example, action counts may differ, or the revenue attributed to exports may not match the platform totals. This is normal and expected, especially when sessions include multiple actions.

Response export reports display data based on when the response occurred, while actions in the platform are shown according to when the spot that triggered the action occurred.

Sessions and actions

A session begins when an end user interacts with an advert and visits a client’s website or app. The session remains active as long as the user continues interacting with the site; it ends after 30 minutes of inactivity.

Within a session, users can perform multiple actions—such as signing up, logging in, or making purchases. They may also repeat the same action multiple times (e.g., several separate purchases). Actions are often referred to as orders, since purchases are typically the most valuable action, which is why there is an order ID column.

Attribution is based on sessions because the session start marks when the customer responds to the advert. Individual actions may occur later, so attributing at the action level could break the connection to the original advert interaction.

The platform reports session counts. When you see 'x number of actions,' it actually represents x sessions that included that action. This approach avoids inflated action counts caused by issues like double-tagging, which can result in duplicate actions within a single session.

Reporting methodologies

A fundamental difference exists between how data is presented in the platform (Pivots) and how it is extracted in raw files (Action Exports).

Pivot

Pivots display actions based on when the advert aired, linking each action to the spot that prompted it. Additionally, the platform reports the number of sessions that contain an action, rather than counting raw actions, to prevent issues such as double-tagging.

Export

Action exports are based on response time and show data from user interactions with the site. Unlike the platform, which aggregates by session, action exports provide details at the individual action level.

Attribution Probability When a session contains multiple actions, reporting becomes more complex. Most reports, such as the platform and spot export, are based on sessions, while the action export provides data at the individual action level.

If a session includes multiple actions, or if a single session is driven by multiple spots and impressions, the session will appear multiple times in the export (once for each action) to display details such as order ID and revenue. This can lead to double-counting, so two attribution probability columns are provided:

1. actionprobability: This column allocates the session's attribution probability across all actions in the session. The values add up to match the session-level attribution reported in the platform. Use this column for aggregate reporting.
2. sessionactionprobability: This column shows the session-level attribution probability, the likelihood that the session was attributable to TV. Since this value is repeated for each action in the session, it should not be summed. Instead, use it for calculating revenue vs. cost and other metrics at the individual action level.

Example

Note:

• The values in the sessionactionprobability and actionprobability columns depend on how many actions (or orders) occur within a session.
• If a session contains only one action, sessionactionprobability and actionprobability will have the same value. If there are multiple actions, such as two orders, the sessionactionprobability remains the same for each action, while the actionprobability is divided among the actions in that session.

Using the data

To match the platform attribution, use the actionprobability column. This value is adjusted to prevent double-counting across multiple actions in a session.

For order-level analysis, refer to the sessionactionprobability column, which reflects the session’s attribution probability for each action.

Note: Keep in mind the differences in timing and time zones: the platform reports the advert time and the pivot time zone, while the export records the response time and the action export time zone. These timestamps usually match closely, but with AdvancedTV or delayed user responses, they can differ by several days.

Glossary of definitions

Related content

Action Export Use Cases