MVP Solution
We wanted to test two variants to learn how asking for all pieces of missing information in a full-screen experience would perform against surfacing prompts for specific missing information one-by-one, every time a user logged in.
Refresh the page if the video isn't playing
Project overview
Collecting missing info on Pinterest
Problem: Pinterest needed to know basic information about users in order to provide a more personalized experience and better targeted ads. The challenge was to figure out how to collect that information for the accounts that already existed without disrupting users too much.
Solution: Together with the team, we’ve created two variants of the design—one is aimed to collect missing information one-by-one after the user logs in and uses the product. Another one is designed to collect all pieces of information in one flow.
Outcome: First name coverage increased by 3%, last name coverage increased by 5%, email coverage increased by 4%, age coverage increased by 2%, gender coverage increase by 2%
More on Process More on Final DesignResearch and brainstorm
After looking at existing data and defining goals for the project, we’ve brainstormed ideas together. A few of them were tight to profile completion but we quickly realized that Pinterest isn't a profile-oriented product and there is no value for users to want to complete their profile.
User testing
We’ve conducted user research to learn how users react to the experience and messaging. We’ve learned that while users didn’t mind being prompted to complete missing information, they wanted more transparency on how the information will be used. It wasn’t really clear how asking for their information like full name would provide a better experience. It was clear that we need to tweak messaging to address those concerns.
Finalized Designs
Outcomes and learnings
First name coverage increased by 3%, last name coverage increased by 5%, email coverage increased by 4%, age coverage increased by 2%, gender coverage increase by 2%
We’ve learned that variant 2 worked better for collecting one piece of missing information, while variant 1 performed better for collecting multiple pieces of missing information. Prompting users to complete information again didn’t perform well.