During UC Berkeley's Information Architecture course, I spearheaded a collaborative case study focused on boosting Craigslist's discoverability and high-revenue task completion.
Team
Noynica Ahuja
Claudia Benitez
Margaret Seymour
Timeline
July 2022 - August 2022
Methodology
Gap Analysis, Task Prioritization, Card Sorting, Tree Testing, Information Architecture
want to get up to speed fast?
What did I do?
Optimized high-revenue tasks and streamlined user journeys from 'Seeking' to 'Doing' by by analyzing the current IA, understanding users' mental models, testing alternate abstract IAs, and ultimately making it easier and faster for users to accomplish tasks on mobile.
What impact did it have?
Tree testing demonstrated a significant improvement in find-ability across the board. Additionally, the introduction of an 'Actions' toolbar aimed at accelerating task completion showed very positive initial effects.
What did I learn?
This project emphasized the value of adapting strategies based on empirical results and user feedback, ensuring a continuous cycle of improvement for digital platforms.
Problem
The rise of social selling on platforms like Facebook has intensified competition for Craigslist. While overall user traffic in 2021 did not fully rebound to pre-pandemic levels, the revenue per transaction has risen. However, many new users perceive the existing organizational system as outdated, crowded, and challenging to navigate. Our emphasis was on improving Craigslist's mobile IA, addressing complexities arising from deep category dependencies and a vast array of items on a small factor form.
Initial Research
Comprised of analyzing stats and trends, and defining user needs and business goals.
business goals
Honor and maintain the values-first simplicity of Craigslist, while updating for more profitable traffic and mobile usage
values
Simplify user connections to be friendly, social, trusting, and communal for both Seekers and Posters.
growth strategy
Organic growth through trusted PEER-to-PEER
cost structure
Maximize community ownership while minimizing overhead through optimized self-regulation.
revenue model
Grow traffic by optimizing usage scenarios (Mobile)
user personas
Next an overview of the target users, their personalities, motivations and pain points was compiled.
task prioritization
Identified a comprehensive list of potential user tasks on Craigslist. To focus our efforts, we highlighted the top tasks:
-
Find a Job / Post a Job Ad
-
Find a Home / Apartment
-
Connect with people sharing common interests
-
Buy / Sell Goods
-
Offer / Get a Service
-
Discover local events
-
Find Gigs / Hire Staff for Gigs
Evaluate and Analyze
gap analysis
A web crawl and gap analysis revealed missed opportunities and performance discrepancies based on identified user tasks.
tree testing
Using Optimal Workshop, the current website format was tested for hierarchy, findability, logic, ease of use, understanding, and navigability by 12 users across India, Canada, and the United States. Out of the eight predefined tasks completed by the testers, only 35% of users reached the correct answer, with 78% of answers chosen without backtracking.
initial promise, navigation woes
observations
-
Unclear selling process : Users landed on "For Sale" despite intent to sell
-
Terminology mishmash : "Jobs," "services," "gigs," and "posts" used interchangeably
-
Community mistaken as forum : Further contributing to confusion.
Most users started correctly, but poor filters and overwhelming options led to wrong subsequent choices.
current challenges
-
Hard to read
-
Scanability issues
-
Poor searchability
-
Exhaustive lists with bad labels
-
Repetitive options and choice overload
improvements
-
Better labeling and grouping of categories
-
Removal of redundant similar categories
-
Introduction of more filters for easier searching
-
Revision of category titles
-
Further work on IA
Synthesize - From Analysis to Sitemap
card sorting
18 participants categorized 40 predefined tasks, labeling similar tasks by group. Labels were then standardized and ranked.
assumption
Users prefer Tasks (nouns) over Categories (verbs)., and communal for both Seekers and Posters.
reality
Most users use nouns & categories, but non-native, mobile & impaired users use verbs alongside nouns.
key insight
Users want to quickly find something and then do something with it.
challenges
Current site map delays the "doing" action.
opportunities
Categories first, then verbs - Guiding users from finding to doing faster.
Verbs as category signifiers - Reduce search time and effort.
This aligns with our goal of helping users getting things faster.
abstract IA and sitemap
Mental model from card sorting and was then used to create an abstract information architecture and the first draft of the new sitemap.
Refine, Validate & Design
label scoring
Labels created in the previous stage after card sorting were evaluated in this phase. All labels with a score of 3 or less are then updated and the abstract IA is amended.
A section of the Scoring template
updated IA
navigation options
Assisted Search
-
Pros - Aligns with user mental model (tree test), reduces confusion from ambiguous terms ("Gigs" vs "Services").
-
Cons: Smaller search box on mobile, potential reliance on suggestions.
Actions Toolbar
-
Pros: Matches user mental model from tree test ("Actions" resonates), especially for verb-based tasks. Solved initial confusion (e.g., "Upgrade a Bicycle").
-
Cons: Requires learning a new interface, needs thorough testing due to potential unfamiliarity.
Navigation options
Updated tree structure on Optimal Workshop
tree testing R2
A second round of tree testing was performed with new abstract IA. 12 participants from across United States, El Salvador and India answered all eight tasks with a success rate of 70%.
round 1
round 2
As seen above, by ensuring focused search and introducing Actions, there’s massive improvement in the findability and IA. To understand better, let’s have a closer look at task 7 (finding a piano class for your brother).
round 1
round 2
Wireframes
The home screen has been redesigned to incorporate the Quick Actions tab. Users can directly jump into the desired sections using this navigation feature.
Results
-
By grouping identical categories into one entity, probability of choice overload is minimized.-
-
Level Two labels in the proposed IA are brief and easy to scan. They will be further split up in Level Three, this ensures easy and time efficient selection and improves readability.
-
Confusion between ‘Gigs’, ‘Jobs’ and ‘Services’ has been reduced to some extent, by creating Hire and Apply in the Actions tool and getting rid of ‘Gigs’ and introducing ‘Freelance’.