Twitter Tourists: Onboarding for Repeat-Churners
Hundreds of millions of users who return to Twitter each month are repeat-churners. Each time they visit, they spend a few days to a week, then disappear for a month or more. Over a few months, I refined our understanding of these users and proposed several exploratory solutions for how we might serve them. These solutions led the team to conduct further research.
Research Findings
We learned a couple things:
Returning users don’t follow or like posts. This makes it difficult to create recommendations.
Returning come back for events and specific accounts.
For this reason, we can’t rely on our usual recommendation signals. Instead, we need to tailor onboarding and recommendations to the events and accounts that bring users back.
Finding 1: Returning users are less-engaged than new users
Compared to new users, they have a low rate of following accounts and liking Tweets. Because of their infrequent visits and transactional, consumption-oriented experience on the platform, I started to call them “Twitter Tourists.”
Returning users like Tweets at a lower rate than new users.
Returning users follow accounts at a lower rate than new users.
Finding Two: Users Don’t Just Return for Events.
The common understanding of repeat visits was that they return for breaking news events and cultural moments. So I asked our data scientist to confirm this with a query of where people visit when they return. What this data showed is that our repeat users visit several different kind of pages. This suggested that they weren’t just coming to Twitter for events.
Events Pages
These visits can be explained by users visiting Twitter for breaking news events and cultural moments. For example, if you come to Twitter to find Tweets about an event, you would go to the explore page to see if it shows up in the top stories of the day. If the event doesn’t show up, you can also search by typing in terms like “earthquake,” “Barbie,” or "election".
Profile Page
The Profile Page visits reflected a pattern that we previously heard about in qualitative research. A user is interested in a celebrity or popular user, so occasionally they’ll return to Twitter to hear what the user has to say before leaving. Yet at the same time, they don’t follow this account because they don’t want this follow to be public. For example, maybe someone is a secret kpop stan, or they’re interested in hearing what a divisive politician is saying but they don’t want to signal support.
Tweet Details
The Tweet details page made sense as a top destination, because this is the page where you see a Tweet and its replies. So regardless of why you’re visiting Twitter, you often end up here. And this was borne out by the data. If you look at where people were coming from, most visits to the Tweet details page come from elsewhere in the app. In summary, visits to the Tweet details page don’t seem to represent a distinct behavior or purpose.
Solutions
Now that we had a clearer picture of who these users are and why they return to Twitter, I started sketching out some design solutions. Since I was still in the early stages of solution exploration, I sketched a few solutions for each problem. These solutions include: more suggestions throughout the search experience, updates inferred by search terms and profile visits, and better privacy controls.
Users Who Return to Twitter for Events
Supportive Search
Problem: After a failed Twitter search, users try a different query or quit altogether.
Solution: Search that provides more suggestions and prevents users from reaching dead ends.
Updates
Problem: Conversations continue even after the user ends a session, and they miss out on updates.
Solution: If a user engages with a trend searches for a term, update them when there has been more conversation related to that trend or search term. These updates can be sent via notification or email.
Users Who Return to Twitter for Tweets from Individuals
Privacy Controls
Problem: Users don’t always want their follows to be public.
Quiet Follow: Quiet follows allow users to follow an account without other accounts being able to see that follow.
Protected Account Prompt: Currently users can make their follows and Tweets private by creating a “Protected Account. If we send this prompt to users when they repeatedly visit an account, perhaps they’ll adopt this new privacy control.
Tweet Updates
Problem: If a user doesn’t follow an account, there’s no way for them to know if an account they’re interested in Tweets. Without knowing, they miss out on updates. Or they log in daily, which is a behavior we’ve heard of anecdotally numerous times.
Infer Interest by Visits: Infer interest without a follow and provide updates via notifications or email.
Impact
This presentation generated interest in the team in learning more about these two segments of users. In particular, the segment of users who visit profiles has previously not been studied, so research will focus on recruiting this group of users and trying to confirm pain points that we’ve heard about anecdotally and identify new ones. More research is planned for later this year.