Academia.edu ・ 2022
Building Academia's first social feed
Turning a document library into a research community
1 Product manager
3 Engineers
1 Data lead
1 Researcher
1 Designer (me!)
Product strategy
Product design
Research & testing
Growth metrics
The challenge
Academia.edu served 20 million researchers but had become just a paper repository. Users arrived from Google, downloaded what they needed, and left immediately. Despite having dense social networks built into the platform-with users averaging 50+ followers each-there was no space for discussion, no way to engage with research, and no community.
The user behavior pattern was clear: researchers checked email 10+ times daily but only visited the site monthly or weekly when they needed a specific paper.
The question: Can we turn Academia into a thriving social network without disrupting our core business?
Built in 4 months with a team of 7
607
monthly feed viewers
17K
discussions created monthly
33K
authors with 100+ followers
26%
of posters posted again
Strategic Constraints
I had to navigate significant business constraints. Single paper pages-where all revenue happened-were completely off-limits for this experiment. Leadership wouldn't risk disrupting our core product.
My solution: use the homepage as a testing ground. It was essentially wasted real estate with low traffic, but that made it safe. We could validate demand without business risk.
The bet: If we could make social work on a page nobody visits, we could scale it everywhere later.
Design Strategy
Email-first engagement
The core design challenge: How do you drive traffic to a page nobody visits?
I turned email into our primary driver. Every engagement-likes, comments, mentions-triggered an email notification that brought users back to the feed. This created a feedback loop:
Email brings user to social feed
User engages with content
Engagement creates visible impact
Impact triggers more emails to others
Cycle continues
Leveraging existing networks
I evaluated two models: a follower model (Twitter-style) versus a relatedness model (Reddit-style). I chose the follower model because academics already had dense networks-when users downloaded a paper, they automatically followed the author.
This meant we could leverage existing relationships and avoid the cold start problem that kills most new social networks. We could surface content immediately without needing to build communities first.
What we built
A clean, academic-focused interface that made sharing ideas as natural as reading a paper.
What we built
A clean, academic-focused interface that made sharing ideas as natural as reading a paper.
The result felt purposeful and academic.
CORE FEATURES
Post Composer
Simple text and link sharing integrated into paper upload flow. Contextual prompts to reduce blank-page anxiety.
Personalized Feed
Posts surfaced to followers and users interested in related topics. Algorithm prioritized recency and relevance.
Engagement Signals
Likes, comments, and email notifications reinforced that someone was listening. Every interaction made the poster feel seen.
Discovery Mechanisms
"Suggested follows" and "trending discussions" helped users find new voices and join conversations.
Key design decisions
Contextual Prompts
The Problem: Generic prompts like "Create a post" left academics unsure what to share, creating friction and hesitation.
My Solution: Context-specific prompts based on where users were and what they were doing - like turning paper uploads into posts, or prompting profile visitors to share research updates.
40% increase in posting rate with contextual prompts vs generic "Create post" button
18% of uploads became posts when integrated into upload flow vs only 2% when separate
Email-First Notifications
The Insight: Academics check email 10+ times daily but visit the site infrequently.
My Approach: Every engagement triggered an email that brought users back to the feed. This made users feel seen and reinforced that their contributions mattered.
65% of returning users came via email
Brand-Aligned Design System
Since we were rebranding during this project, I designed all social components from scratch to match the new brand. This meant creating an entire design system-every feed card, composer, and button-optimized for social interaction while establishing the new Academia identity.
I created and owned this design system simultaneously while testing and iterating on the MVP.
Social integration
This wasn't just a feed in isolation - it was a system that connected our highest-traffic pages and funneled users into our core product.
How interactions flowed:
User sees a post from an author they find interesting → clicks through to their academic profile page
User discovers a paper in the feed → clicks into the single paper page to read and engage with the research
User clicks on a topic tag → navigates to our topic pages to explore more papers and discussions in that research area
I updated our topic pages with the new brand as part of this work, ensuring a cohesive experience as users moved between the feed and other core pages.
Launch and early results
We launched quietly to tens of thousands of academics. Within the first month, we had clear signal this was working.
WHAT WORKED
607
users (2.8%) posted at least once
26%
of first-time posters posted again
53%
of readers viewed multiple posts
~10K
unique readers per week
22%
of posts earned 100+ views
12%
of posts received 10+ likes
Key Learnings
Network effects equal feedback loops
If people can't see their impact, they leave. Every like, comment, and notification had to feel meaningful and pull users back in.
Optimize for your power users first
The 2.8% of posters created all the value-I designed to maximize their activation and retention.
Test activation paths early
Finding the right prompts and entry points made the difference between 2% and 18% conversion.
Start where it's safe, scale where it matters
Proving value on low-risk pages earned us permission to transform the core product.
