To significantly improve ad relevance on X, we've identified three key optimizations for the Custom Audience product:
Build audiences based on your first-party data like customer lists, website visitors, and app users, or users IDs.
In the study, we used Custom Audiences created with public user IDs gathered from the X API. We used Socialdatabase's product 'X Pro Targeting' to populate the Custom Audiences.
In all cases, Custom Audiences built with public X data has outperformed all other targeting features, scoring 86% in ad relevance on average with the highest score of 96%.
Target audiences by including or excluding keywords and related terms they've searched, Tweeted, or from Tweets they've engaged with.
We used keywords like 'tennis' to reach tennis fans, and keywors like 'oncologist' to reach oncologists.
We expected this feature to perform well, but it showed an average relevance of 21% for hashtags and 17% for keywords. This is likely due to the fact that it includes users who have engaged with posts that contained such keywords, scaling the audience significantly beyond the actual usage of keywords.
Reach people with similar interests to an account's followers. Similarity is based on things like Tweets, Retweets, clicks, and more.
We used very relevant follower handles, accounts with followers that have a high density of the audience in their follower based. We tested narrow handles (the most relevant followers) and broad handles (the most scalable accounts within a topic).
Follower look-a-like targeting showed an average relevance of 52% when using narrow handles, and 30% when using broad handles.
Allow X to reach people beyond your targeted audience when there's potential to achieve better performance. This feature is always on for X Premium Users.
We used optimize targeting both in isolation and in combination with other targeting features. We thought that when using it alongside relevant targeting, it would be able to better predict what other users could be relevant. We also created seperate campaigns such as: 'Do you like Tennis?" with yes, no, and I don't know options. To see if optimize targeting can optimize when a single important keyword is used.
Optimize targeting was in all cases the least performing feature, with an average relevance of 6%. It seems to scale your audience at a low CPM, but at the cost of ad relevance.
It’s OBVIOUS that X offers the best value for money. It’s also clear that X is leaving billions in revenue on the table - here’s how it will be unlocked.
— Thomas Slabbers (@Thomasslabbers) December 2, 2024
The most influential people in the world are active on X, and you can reach them for peanuts.
It all comes down to ad… pic.twitter.com/pjrHbQyK2r
When Elon talked about ‘populating a vector space’ to improve ad relevance, I noticed many confused faces in the audience at the Cannes Lions. Here’s what it means:
— Thomas Slabbers (@Thomasslabbers) June 20, 2024
In machine learning, a vector space is a mathematical space where entities (like words, images, videos) are… pic.twitter.com/S92bFpBP72
Several of our advertisers noticed that while running campaigns on both X and Meta, nearly all conversions are attributed to Meta, with X reporting almost no conversions. Interestingly, when X Ads are turned off, total conversions significantly drop, including those reported on… pic.twitter.com/ChVVSXAoVS
— Thomas Slabbers (@Thomasslabbers) May 11, 2024
Mysocialdatabase.com B.V.
Slego 1a
1046 BM Amsterdam
KvK number: 58951326