The problem
Tumblr’s onboarding could take a new user through 20 to 25 screens before they reached their dashboard, and most of them never came back for a second session.
We ran usability tests with non-users aged 18–25 and ran research alongside Irregular Labs focused on Gen Z new users who’d been on the platform two to six months. We also audited the existing flow ourselves, going through it step by step on mobile. Two problems kept coming up across every source.
Topic selection felt like a broken promise. New users took it seriously. They spent real time picking topics they cared about, but the feed they landed on afterward didn’t reflect those choices. The usability research called the onboarding experience “rife with unfulfilled promises which dissuade people who are on the fence about using Tumblr regularly.” People liked the idea of Tumblr but weren’t getting to the content they came for fast enough.
Login and signup were separate flows. You had to decide upfront whether you had an account. Pick the wrong one and you’d have to back out and start over.
Fig 01 The existing iOS onboarding flow.
From topics to blogs
Fig 02 The topic selection area. Good information density, but static, content-free, and not personalized enough.
The Irregular Labs research added a layer to the first problem. What kept new users around wasn’t just finding content. It was having an early engagement with another person, through a like, a reblog, or an ask. Topic selection alone couldn’t get someone there because topics are abstract. You don’t engage with a topic. You engage with a blog.
The engineering team’s assessment was that topic onboarding had hit its ceiling. The existing picker was static and hand-curated, didn’t show any taste of what you’d actually see, and the supertopics were so general they didn’t lead to interesting enough content. If topic selection had hit its ceiling, the next lever was getting users to follow actual blogs: people and communities, not just broad categories.
Exploring flows
I stripped the flow down to steps and goals to understand where we could improve.
Fig 03 Mapping possible flows against user goals and technical constraints.
Fig 04 Shortened flow. Prioritize finding content, get administrative info out of the way quickly. This is the direction we went with.
Fig 05 Gamifying the exploration to encourage interactions before account creation.
We liked the idea of letting users explore content before creating an account. Reddit’s app does this well. But the research had told us the retention signal was an early engagement with another person, and persisting those interactions without an account to link them to was a bigger lift than we wanted to take on first. We went with the shortened flow and focused on getting users to follow real blogs faster.
What we shipped
The redesign addressed both problems directly. Login and signup were combined into a single flow where you enter your email and the system determines whether you’re new or returning. Sub-topics were visible by default, updating dynamically so the editorial team didn’t have to curate by hand.
The biggest change: blog recommendations were added to onboarding alongside topics, so new users would follow real blogs with recent posts, not just abstract tags. If your feed is populated by blogs you actively chose, your first session is more likely to contain something worth engaging with. That’s the path the research had identified to that first connection.
Fig 06 The shipped onboarding flow.
Results
We A/B tested on 50% of iOS traffic over a month. Blog following surged, as expected (the step was brand new). The real signal was what happened after:
- 28-day retention
- ~2x
Most new users never returned for a second session, so nearly doubling 28-day retention confirmed the approach worked. We scaled it to mobile web and desktop.