DESIGNWNOY
Revamping Craigslist Mobile for the “Getting Things Done” Generation
Revamping Craigslist Mobile for the “Getting Things Done” Generation
During UC Berkeley's Information Architecture course, I spearheaded an IA overhaul to improve usability and support Craoglist's mission of community-driven growth and high-revenue task completion.
During UC Berkeley's Information Architecture course, I spearheaded an IA overhaul to improve usability and support Craoglist's mission of community-driven growth and high-revenue task completion.
Team
Team
Noynica Ahuja, Claudia Benitez, Margaret Seymour
Noynica Ahuja, Claudia Benitez, Margaret Seymour
Timeline
Timeline
July 2022 - September 2022
July 2022 - September 2022
Methodology
Methodology
Gap Analysis, Task Prioritization, Card Sorting, Tree Testing, IA
Gap Analysis, Task Prioritization, Card Sorting, Tree Testing, IA
(TLDR)
(TLDR)
Craigslist’s mobile maze was losing users until we
Craigslist’s mobile maze was losing users until we
restructured journeys from seeking to doing
restructured journeys from seeking to doing
The Problem
The Problem
Craigslist’s trademark minimalism had become a liability on mobile. With hundreds of nested categories squeezed onto small screens, users spent more time hunting than completing tasks.
Craigslist’s trademark minimalism had become a liability on mobile. With hundreds of nested categories squeezed onto small screens, users spent more time hunting than completing tasks.
My Role
My Role
Led research and design activities including tree testing, card sorting, IA synthesis, and wireframing, collaborating with two peers.
Led research and design activities including tree testing, card sorting, IA synthesis, and wireframing, collaborating with two peers.
Impact
Impact
Restructured Craigslist’s mobile IA from static categories to action-oriented flows, preserving its minimalist ethos while enabling faster task completion. This shift surfaced high-value actions earlier and improved direct success rates by 35–50% across key tasks.
Restructured Craigslist’s mobile IA from static categories to action-oriented flows, preserving its minimalist ethos while enabling faster task completion. This shift surfaced high-value actions earlier and improved direct success rates by 35–50% across key tasks.
(The Problem)
(The Problem)
Craigslist’s simplicity was breaking
Craigslist’s simplicity was
under its own weight
breaking under its own weight
A maze for everyday tasks
A maze for everyday tasks
Once the go-to marketplace for local communities, Craigslist built its reputation on simplicity and trust. But a closer look at its mobile information architecture revealed critical gaps:
Hundreds of nested categories on a small screen
High-friction task paths and unclear revenue-driving flows
Vague, overlapping labels
Once the go-to marketplace for local communities, Craigslist built its reputation on simplicity and trust. But a closer look at its mobile information architecture revealed critical gaps:
Hundreds of nested categories on a small screen
High-friction task paths and unclear revenue-driving flows
Vague, overlapping labels
Users often began confidently but quickly stalled; our first round of testing showed only about one in three tasks ended in success.
Users often began confidently but quickly stalled; our first round of testing showed only about one in three tasks ended in success.
Business at a tipping point
Business at a tipping point
Each failed task erodes trust, slows down peer-to-peer exchanges, and nudges frustrated users toward faster-moving rivals like Facebook Marketplace and OfferUp.
Each failed task erodes trust, slows down peer-to-peer exchanges, and nudges frustrated users toward faster-moving rivals like Facebook Marketplace and OfferUp.
Design Statment
Design Statment
How might we reduce category clutter and decision fatigue on Craigslist mobile while preserving its no-frills, community-driven ethos?
How might we reduce category clutter and decision fatigue on Craigslist mobile while preserving its no-frills, community-driven ethos?
(Research Methods Applied)
(Research Methods Applied)
Finding where users got lost
Finding where users got lost
Before I mapped the cracks in Craigslist’s IA, I stepped back to define the larger-scale goals at play. For Craigslist, growth meant optimizing high-value mobile tasks while preserving its low-overhead, community-first model. For users, it meant being able to trust the platform to help them complete everyday tasks, without confusion or wasted effort.
Before I mapped the cracks in Craigslist’s IA, I stepped back to define the larger-scale goals at play. For Craigslist, growth meant optimizing high-value mobile tasks while preserving its low-overhead, community-first model. For users, it meant being able to trust the platform to help them complete everyday tasks, without confusion or wasted effort.
Zeroing in high-impact tasks
Zeroing in high-impact tasks
Craigslist supports dozens of actions, but not all carry equal value and redesigning every edge case wasn’t realistic, or strategic.
From 25 possible tasks, I prioritized those most in demand and most tied to revenue. A task inventory and impact analysis surfaced the core of Craigslist’s peer-to-peer engine: jobs, housing, goods, services, events, and gigs. These high-value actions became the North Star for evaluating the IA.
Craigslist supports dozens of actions, but not all carry equal value and redesigning every edge case wasn’t realistic, or strategic.
From 25 possible tasks, I prioritized those most in demand and most tied to revenue. A task inventory and impact analysis surfaced the core of Craigslist’s peer-to-peer engine: jobs, housing, goods, services, events, and gigs. These high-value actions became the North Star for evaluating the IA.
Stress-testing existing IA
Stress-testing existing IA
With high-impact tasks defined, I ran a gap analysis followed by tree testing to see how well the current IA supported them.
With high-impact tasks defined, I ran a gap analysis followed by tree testing to see how well the current IA supported them.
Gap analysis showed structural cracks
Gap analysis showed structural cracks
Using PowerMapper Cloud, I crawled the site (499 pages) and mapped them against existing categories. This exposed three major issues:
Missed opportunities — tasks with no clear path
Broken mental models — flows misaligned with user expectations
Redundant categories — overlapping labels causing friction
Using PowerMapper Cloud, I crawled the site (499 pages) and mapped them against existing categories. This exposed three major issues:
Missed opportunities — tasks with no clear path
Broken mental models — flows misaligned with user expectations
Redundant categories — overlapping labels causing friction
The analysis showed that high-value tasks were buried under clutter or forced into confusing paths.
The analysis showed that high-value tasks were buried under clutter or forced into confusing paths.
Users lost their way after the first click
Users lost their way after the first click
Next, I led the tree test with 12 participants across India, Canada, and the U.S.
Success rates were only 35%. Most users started in the right place, vague labels, weak filters, and long lists led them astray. Posting jobs, selling items, and navigating between Community and Forums were frequent failure points.
Next, I led the tree test with 12 participants across India, Canada, and the U.S.
Success rates were only 35%. Most users started in the right place, vague labels, weak filters, and long lists led them astray. Posting jobs, selling items, and navigating between Community and Forums were frequent failure points.
The problem wasn’t discovery, it was progression
The problem wasn’t discovery, it was progression
Users don’t come to Craigslist to browse endlessly. They come to get something done - post, sell, apply, find. The current IA slows them down instead of helping them act.
Users don’t come to Craigslist to browse endlessly. They come to get something done - post, sell, apply, find. The current IA slows them down instead of helping them act.
Unclear Workflows
Unclear Workflows
Creating a listing was split between For Sale and Create a Posting, leaving users unsure how to start.
Creating a listing was split between For Sale and Create a Posting, leaving users unsure how to start.
Confusing Labels
Confusing Labels
Categories like Jobs, Services, Gigs, and Community overlap. Users hesitate, backtrack, or abandon tasks altogether.
Categories like Jobs, Services, Gigs, and Community overlap. Users hesitate, backtrack, or abandon tasks altogether.
Weak filters, overloaded lists
Weak filters, overloaded lists
Even after reaching the right category, endless unfiltered lists cause fatigue and decision paralysis.
Even after reaching the right category, endless unfiltered lists cause fatigue and decision paralysis.
(From analysis to a clearer sitemap)
(From analysis to a clearer sitemap)
Understanding mental models through card sorts
Understanding mental models through card sorts
Nouns orient, Verbs guide
Nouns orient, Verbs guide
I ran 18 card sorts (online and in-person) with participants grouping 40+ Craigslist tasks into categories that felt natural to them. To avoid bias, I normalized across three native-language speakers (English, Spanish, Hindi). This surfaced the most intuitive, consistent labels across linguistic and cultural differences.
I ran 18 card sorts (online and in-person) with participants grouping 40+ Craigslist tasks into categories that felt natural to them. To avoid bias, I normalized across three native-language speakers (English, Spanish, Hindi). This surfaced the most intuitive, consistent labels across linguistic and cultural differences.
Current IA felt familiar but trapped users in “finding” mode
Current IA felt familiar but trapped users in “finding” mode
when what they really wanted was to find → then do
when what they really wanted was to find → then do
What Worked
What Worked
People consistently grouped around broad nouns like Jobs, Housing, Services. These categories felt familiar and provided a mental “home” for different tasks.
People consistently grouped around broad nouns like Jobs, Housing, Services. These categories felt familiar and provided a mental “home” for different tasks.
What was Confusing
What was Confusing
Actions like Sell, Post, Apply, and Find overlapped across categories, leaving users unsure where to go. Sell under For Sale or Post? Gigs under Jobs or Services? This led to backtracking, extra clicks, or abandonment.
Actions like Sell, Post, Apply, and Find overlapped across categories, leaving users unsure where to go. Sell under For Sale or Post? Gigs under Jobs or Services? This led to backtracking, extra clicks, or abandonment.
Interestingly, non-native speakers and mobile-first users preferred verb–noun phrasing like Sell Item or Find Gig. For them, pairing the action with the object improved comprehension and reduced hesitation.
Interestingly, non-native speakers and mobile-first users preferred verb–noun phrasing like Sell Item or Find Gig. For them, pairing the action with the object improved comprehension and reduced hesitation.
Consolidating Categories from Card Sort
Consolidating Categories from Card Sort
From the card sort, I normalized labels and combined overlapping areas. This produced a clearer, user-driven set of high-level categories. For Sale, Jobs, Services, Community, Housing, Activities, Entertainment, Freelance.
From the card sort, I normalized labels and combined overlapping areas. This produced a clearer, user-driven set of high-level categories. For Sale, Jobs, Services, Community, Housing, Activities, Entertainment, Freelance.
These noun-based anchors reflected how people actually thought about Craigslist, familiar enough to orient, but simplified to reduce clutter and duplication.
These noun-based anchors reflected how people actually thought about Craigslist, familiar enough to orient, but simplified to reduce clutter and duplication.
Flows Reimagined as Quick Actions
Flows Reimagined as Quick Actions
With the high-level categories consolidated, the next challenge was reducing friction within them. The card sort had revealed that while nouns gave users orientation, it was verbs that guided their momentum.
With the high-level categories consolidated, the next challenge was reducing friction within them. The card sort had revealed that while nouns gave users orientation, it was verbs that guided their momentum.
To translate that insight into design, I mapped user flows across each new category and identified the most common actions:
Jobs (incl. Freelance, Resumes) → Apply, Hire
Marketplace (For Sale) → Buy, Sell
Housing → Rent
Services, Community, Activities, Entertainment → Connect
Instead of burying these verbs under multiple menus (Sell inside For Sale, Apply inside Jobs), I elevated them into a Quick Actions tab.
To translate that insight into design, I mapped user flows across each new category and identified the most common actions:
Jobs (incl. Freelance, Resumes) → Apply, Hire
Marketplace (For Sale) → Buy, Sell
Housing → Rent
Services, Community, Activities, Entertainment → Connect
Instead of burying these verbs under multiple menus (Sell inside For Sale, Apply inside Jobs), I elevated them into a Quick Actions tab.
This created a dual-entry system that supported two mental models:
Categories (nouns): Jobs, Housing, Marketplace, Services, Community, Activities, Entertainment, Freelance.
Quick Actions (verbs): Buy, Sell, Rent, Apply, Hire, Connect.
By grounding Quick Actions in both user mental models (card sort) and business priorities (task analysis), the IA shifted from overwhelming lists into action-oriented funnels. Users could now either browse broadly by category or jump straight into the task they came for, reducing friction and wasted clicks.
This created a dual-entry system that supported two mental models:
Categories (nouns): Jobs, Housing, Marketplace, Services, Community, Activities, Entertainment, Freelance.
Quick Actions (verbs): Buy, Sell, Rent, Apply, Hire, Connect.
By grounding Quick Actions in both user mental models (card sort) and business priorities (task analysis), the IA shifted from overwhelming lists into action-oriented funnels. Users could now either browse broadly by category or jump straight into the task they came for, reducing friction and wasted clicks.
Pruning with label scoring
Pruning with label scoring
Next, I ran label scoring (clarity, recognition, distinctiveness) and a similarity matrix to clean up the structure:
Next, I ran label scoring (clarity, recognition, distinctiveness) and a similarity matrix to clean up the structure:
Removed overlaps → Gigs became Freelance; Discussion Forums folded into Community.
Clarified vague terms → For Sale reframed as Marketplace; Jobs absorbed Resumes.
Flattened long lists → similar subcategories collapsed, with details moved into filters (price, condition, location, etc.).
Removed overlaps → Gigs became Freelance; Discussion Forums folded into Community.
Clarified vague terms → For Sale reframed as Marketplace; Jobs absorbed Resumes.
Flattened long lists → similar subcategories collapsed, with details moved into filters (price, condition, location, etc.).
Bringing it all together into a sitemap
Bringing it all together into a sitemap
The restructured sitemap brought together everything learned from card sorting, tree testing, and label scoring.
Quick Actions (verbs) gave users shortcuts to core tasks: Buy, Sell, Rent, Apply, Hire, Connect, Donate.
Main Menu (nouns) provided clean anchors: Jobs, Housing, Marketplace, Services, Community
Subcategories were flattened and long lists replaced with filters (price, location, condition), cutting down choice overload.
The restructured sitemap brought together everything learned from card sorting, tree testing, and label scoring.
Quick Actions (verbs) gave users shortcuts to core tasks: Buy, Sell, Rent, Apply, Hire, Connect, Donate.
Main Menu (nouns) provided clean anchors: Jobs, Housing, Marketplace, Services, Community
Subcategories were flattened and long lists replaced with filters (price, location, condition), cutting down choice overload.
(Validation and Further Refinements)
(Validation and Further Refinements)
Tree Test Round 02
Tree Test Round 02
Armed with a flattened sitemap and Quick Actions, we ran another round of tree testing with 12 participants across the U.S., El Salvador, and India.
Success rate jumped to 80%.
Directness increased - Quick Actions allowed users to act immediately without scanning long lists.
Fewer backtracks and faster time-on-task were observed.
Armed with a flattened sitemap and Quick Actions, we ran another round of tree testing with 12 participants across the U.S., El Salvador, and India.
Success rate jumped to 80%.
Directness increased - Quick Actions allowed users to act immediately without scanning long lists.
Fewer backtracks and faster time-on-task were observed.
Key wins:
“Sell an item” became direct via the Sell Quick Action instead of hidden in “For Sale.”
“Apply for a job” flowed naturally under Jobs → Apply instead of bouncing between Jobs, Gigs, and Services.
Key wins:
“Sell an item” became direct via the Sell Quick Action instead of hidden in “For Sale.”
“Apply for a job” flowed naturally under Jobs → Apply instead of bouncing between Jobs, Gigs, and Services.
To understand better, let’s have a closer look at task 7 (finding a piano class for your brother). In Round 1, most participants wandered between Community and Services without success. After the redesign, most users went directly to Services → Lessons & Tutoring, completing the task quickly and with far less hesitation.
To understand better, let’s have a closer look at task 7 (finding a piano class for your brother). In Round 1, most participants wandered between Community and Services without success. After the redesign, most users went directly to Services → Lessons & Tutoring, completing the task quickly and with far less hesitation.
Despite the improvements, some friction remained. When asked to sell an item, several users still clicked into Marketplace instead of using the Quick Action Sell → Post an Ad. This revealed that while Quick Actions helped, Marketplace still signaled “the place for items,” causing hesitation.
Despite the improvements, some friction remained. When asked to sell an item, several users still clicked into Marketplace instead of using the Quick Action Sell → Post an Ad. This revealed that while Quick Actions helped, Marketplace still signaled “the place for items,” causing hesitation.
Final Refinements
Final Refinements
To address these sticking points, I made a final set of changes to the IA before moving into wireframes:
Sell clarified: Marketplace locked to browsing; Quick Action relabeled as Post an Ad to reduce ambiguity.
Freelance integrated: Nested under Jobs rather than standing alone, keeping the IA predictable and shallow.
Community simplified: Forums, Lost & Found, and Groups consolidated under Community to reduce overlap.
Filters standardized: Applied consistently across Housing, Marketplace, and Jobs for easier scanning.
To address these sticking points, I made a final set of changes to the IA before moving into wireframes:
Sell clarified: Marketplace locked to browsing; Quick Action relabeled as Post an Ad to reduce ambiguity.
Freelance integrated: Nested under Jobs rather than standing alone, keeping the IA predictable and shallow.
Community simplified: Forums, Lost & Found, and Groups consolidated under Community to reduce overlap.
Filters standardized: Applied consistently across Housing, Marketplace, and Jobs for easier scanning.