BACK
ELKE DIANE GENTILE
AD PREFERENCE
CONTROLS
THE PROBLEM
People using Facebook have a sense that their actions are being tracked in order to change their algorithms but they are not in the driver's seat. This leads to behaviors that send mixed signals to ranking and ends up producing a less relevant browsing experience.
There was also a concurrent issue with regulations requiring an opt in model for any sort of usage tracking so the need for explicit responsive controls was not just a single pronged issue.
THE PLAN
Alignment
Ads exist app wide so the need to cross functional alignment between every surface team, as well as with the larger design systems team, was step one.
I used a stepped approach to implementation with the focus being on high traffic areas first to refine a set of guidelines for implementation elsewhere.
Communication
Making sure teams stayed up on the progress and thinking was a large operational challenge.
In order to keep myself from having to attend endless briefing meetings, I created a microsite that outlined the reasons for the work (citing research studies we had conducted) as well as links to ongoing efforts in the project.
Partnership
I had 2 teams that I worked with very closely for this effort. First was the integrity team, which was also handling similar controls for non ad content. Second was the larger Meta ads platform which sat above all the apps and would be the layer where any feedback given could be reviewed and edited by the end user.
THE PROGRESS
Coming into this work, there were a few one off patterns that had been tested in order to help build learnings.
The issue was these were slated to be deprecated because they didn't work broadly enough. I was tasked with evolving what was promising data into widely adoptable patterns.
This was rather challenging because there was an explicit request to not conflate it with the existing feedback reactions.
I ended up working closely with the organic team on this so we could more quickly reach alignment for test candidates that could actually ship.
By the end, we had created a fairly robust modular system that we took into qualitative research so we could narrow down candidates. I created prototypes and worked with the external research vendors to craft prompts and tasks for users to go through.
THE PRODUCT
After much back and fourth about the degrees of control people said they wanted vs what we were seeing in usage we ended up actually going back to a part of the patterns that initially were considered "un-shippable" because they were thought to be too destructive to organic content recommendation. We also kept options for more nuanced feedback but simplified those experiences as well.
THE POST LAUNCH
The main success metrics for this product were first, don't regress any time spent on the platform, and secondly engagement with the feature itself. The effort was successful in both arenas and we were able to increase engagement to 5% (which for context was pretty big given usual metric movement is 1% or less).
This project was also a really shining example of what people say they want vs what they will actually use can be a pretty different. It was a great experience to have so much research and test data to pull from in order to push product decisions up the ladder instead of taking only top down direction.
2026 ELKE DIANE GENTILE
CONTACT
RESUMÉ