xandr Assignment Flow - The Segment Manual

Tags

Top Highlights

• However these updated assignments are not immediately propagated to xandr. xandr assignments are made on an hourly basis based on the users that were active in the previous hour. The active users from the previous hour is found based on a dataset generated by Pulse that streams the user identifiers of events that happen during an hour into a dataset on S3. This streaming to S3 means that the data is available almost immediately, at most with a 5 minute delay unless there are other delays in the Pulse streaming pipeline.

• For a user to be assigned to a segment the Segment system needs to receive data about the user that matches the criteria of that segment

• When the Segment system has consumed this data, either through the Import Predictions, Import Lookalikes or Import Custom Audience tasks, the next matching job that runs will pick up this new data and match the user to the segment, updating that user's assignments in the user_segments table.

• When these criteria are met the Exports job will pick up the assignments for that user from the user_segments table, associate the given advertisement identifier with those segments, and upload this assignment to xandr.

• When generating the xandr file this dataset is read and the last event for each advertisement identifier is located. The user identifier of this event is then used to look up what segments this advertisement identifier should be assigned to. This means that the requirement for a user being targetable for an advertisement using a given segment is not simply that the data for matching that segment's criteria is available to Segment, but rather the following has to happen in order: The data for matching the segment's criteria is imported into Segment An event must happen that includes that user's identifier along with an advertisement identifier The user's identifier must be the last user identifier associated with the given advertisement identifier within that hour