Transparency in Top Friend Suggestions

Skills I used:

  • A/B testing

  • Content and visual consistency/scalability

  • Transparency

Summary

Your first launch of any feature or experience usually isn’t the last. I’ll walk you through some of the ways I improved the content and design of the Top Friend Suggestions feature at Facebook and their impact to core growth metrics.

The Problem

With the recent launch of a prominent friend suggestion unit in the People You May Know (PYMK) section, users had complained that it:

  1. Took up a lot of space.

  2. Didn’t know why a certain person was highlighted over another.

  3. Didn’t give them an option to hide the unit.

Without a clear name and clear value prop that provided little user agency, this PYMK unit had clear quality issues.

The Solution

Our product team ran a series of three tests to address the above people problems to improve the quality of the unit. The tests we ran were:

  • A social context test, to understand the value of each piece of social context by testing short vs. full names, bolding and the number of top mutual friend names shown.

  • A hide test, which tested 2 overflow menu designs and various hide durations.

    • I got buy-in from the team to test an added layer of transparency into the unit, which would explain to them why the unit is relevant by providing them with a "Why am I seeing this?" section in the overflow menu.

  • A height test to compare different heights and UI layouts to reduce the space taken on the screen without hurting core friending metrics.

We found through experimentation that the far right card (237 px) performed the best. The three dots allowed us to provide other actions than just Add and Remove, like the Hide and Report actions later on. It also helped spread out the mutual friends text, compared to how crammed and visually overwhelming the text appears in the far left iteration with the shortest card height.

In later iterations, the designer and I decided to center the profile picture to create more visual differences between a regular friend suggestions and a top friend suggestion. We also pulled the top friend suggestion’s Facebook cover image (if there was and if their privacy settings allowed it), to drive even more visual difference.

For the “explanation” variant, I structured the content and design similarly to the Facebook Privacy team and had their CD team review it in Privacy Office Hours to maintain scalability and visual consistency throughout the Facebook app.

One drawback of this iteration was that we didn’t have room to provide the time limit up front, that this module would only be hidden for the next 24 hours. If someone selected “Hide Top Suggestions,” they’d get a confirmation toast outlining that the module would be hidden for 24 hours.

This was a tradeoff I wanted to make because I believed that providing “the why" would reduce the amount of Hides vs. the variant that was purely information and functional in its approach that provided that detail up front.

The Results

After experimenting, we found that the added explanation as to why they were seeing the unit (labeled “1 day (explanatory)” in the chart) led to a high conversation rate (0.8824%), nearly as high as the control (0.8892%).

My hypothesis about less hides was also correct: when users tapped into the overflow menu, they were less likely to hide the feature in the variant that provided the why (25.07%) versus the one that directly gave the time limit information up front (29.7%).

This could be an indication that the explanatory text did help users understand value of the unit.

The team aligned on shipping the hide variant that included the “Why am I seeing this?” text because it produced the least amount of hides and the closest conversion rate to the control, which didn’t have the hide option.

The Impact

We launched the quality improvements with a card height or 237 px, shipping the hide design with explanatory content, and a hide duration of 7 days instead of just 24 hours after discussion with the product team and initial user feedback in research tests. (The content would inform the user about the 7 day time limit if they selected “Hide Top Suggestions” CTA.)

These quality content and design improvements led to:

  • +11k friend requests sent from unit in the U.S.

  • +0.85% friend suggestion impressions in the U.S.

  • No significant drops in our topline metrics.

The main takeaway from this project? Transparency is key.

The Retro

Looking back, I don’t think we solved the first user problem, which was that the module was taking up a lot of space. Though I do think with later iterations, where we centered the image and showed the user’s profile cover image when possible, I think we made the design visually different enough to justify the longer height.