About offline event data deduplication

30/11/2019

When you upload offline event data, duplicates of events that were already uploaded to that offline event set in the last four days will automatically be skipped or deduplicated. This process helps ensure accuracy in your offline event and attribution reporting.

How it works

When an offline event data file is being processed for upload, Facebook's system will detect and skip duplicates of events that were already matched and uploaded in the last four days. The combination of data used for detecting duplicates is: Offline event set ID + event timestamp + hash of match keys + event name + order ID + item number (optional).

This process is helpful for when:

  • You upload daily at different times and there is some overlap with the previous day's upload.
  • You want to re-upload an event file with additional or updated customer data fields, or match keys, to try to increase the number of matched events. Match keys are only used for matching events to people to help determine which events to upload, and are not deduplicated.
  • You have partially-completed uploads across multiple attempted uploads.

Note: The deduplication process is designed to compare new event data against previously matched and uploaded events from the last four days in order to detect and upload new events only; the process is not intended to edit or update previously uploaded event values.

Best practices for preparing your data

If the same person makes multiple purchases at different times, you can better ensure that all purchase events are processed if you upload the precise timestamps.

However, only the first event will be processed (and the rest skipped) if:

  • You upload a single purchase as multiple events, such as splitting each line item into its own event without providing order_id and item_number.
  • You don't have precise timestamps, such as if your timestamps only display dates (no hour, minutes, seconds), and a customer makes separate purchases on the same day.

To make sure that all unique purchase events are processed, you can provide order_id (unique transaction IDs) and item_number so Facebook can identify the events as unique and not duplicates.

Note: Offline conversions don't currently use the order_id field for order ID reporting; it would only be used to prevent inaccurate deduplication as described for the scenarios above.

* Nguồn: Facebook