About measuring sales impact

How we measure the effectiveness of posts from brands


We measure the performance of brands' posts (both Promoted and organic) to help them provide high-quality ads and content that is more interesting to you. We may work with ad partners to help measure the impact of such ads and content based on information they have about in-store sales. We hope that this helps increase the usefulness of brands' posts for you.

Let's say a food brand would like to know how its posts impact sales of its ice cream in stores. We've designed a process to measure this offline sales impact while protecting user privacy. Here's how it works:

Our measurement partners send us a list of browser cookies IDs and/or unreadable scrambled (hashed) email addresses, mobile device IDs, and phone numbers, along with corresponding ID numbers that we then match to corresponding information that our users have associated with their X accounts. When there's a match, we share with our measurement partner only the matched ID number and information about activity on X, like views of and clicks on the food brand's posts. Then our partner can measure the offline sales impact by comparing the ice cream purchases of users shown the brand's content on X with those who were not. Our partner includes the aggregated results of its analysis in a report for the brand.

 

Measurement with privacy in mind


We have taken extra steps to respect privacy in our measurement process, including hiring a top-tier consulting firm to perform an independent review of the privacy practices of Datalogix, the first partner that X worked with to measure offline sales impact. The firm confirmed that:

  • Datalogix never sends X information about individual users' purchases.
  • The reports that Datalogix creates for advertisers include only aggregated information.
  • The reports have minimum group sizes that are large enough that they do not reveal individual users' purchases.
  • Datalogix segregates the information it receives from X and does not incorporate it into its existing data sets about individual users.

Share this article