Rainbow Sheet
I compared recorded behaviors across participants to identify recurring patterns and differences between sessions.
Usability testing research · Product discovery
- Main Role
- User Research Lead
User-pattern analysis - Duration
- January – March 2026
- Project Type
- Academic Usability Research
- Tools
- Figma · Maze
Project Overview
Understanding how shoppers find the right item.
This academic usability testing research project explored how to improve the experience of finding a suitable secondhand item on thredUP. We observed how shoppers searched, filtered, and evaluated products against specific needs, using Depop as a comparison. Moderated usability sessions and two Maze prototype test rounds helped us identify friction, analyze user patterns, and develop research-informed recommendations. Filter discovery and size understanding emerged as key areas for further exploration.
Making user patterns visible.
I analyzed behavior across sessions to connect individual observations with recurring patterns.
Affinitization
I grouped related observations into themes, then worked with my teammates to interpret the findings and prioritize what to test next.
I also moderated approximately one-third of the usability sessions and built and iterated the Maze prototype. Synthesis, prioritization, and final recommendations were shared team work.
See how observations became patterns ↓Explore this case study
01 / Study framing
Start with a testable question.
Our review of thredUP and Depop helped frame the user tests: where do shoppers hesitate while finding a suitable item?
We reviewed four workflows, then focused testing on product discovery and the steps toward purchase.
Why Depop?
fashion
SIMILARITIES
Comparable shopping tasks
BrowseEvaluateSavePurchase
Shared interest in affordable, sustainable fashion.
Differences
DIMENSIONthredUPDepop
Platform model
Curated resale
Peer-to-peer marketplace
Content
Standardized product information
Seller-generated listings
Purchase decisions
Fixed prices
Offers & negotiation
Trust
Platform-level trust
Seller profiles & social signals
RESEARCH VALUECompare how different interfaces support the same shopping tasks—and where uncertainty appears.
Workflow Scope
Four workflows reviewed across both platforms
- Purchase
- Product discovery
- Profile
- Favorites
Purchase
Cart → Checkout
Ambiguous cart add-on
Cart page · Item list
1Optional bag resembles a cart item.
Shipping cost deferred
Shopping bag · Order summary
2Total excludes the unknown shipping cost.
Product discovery
Search → Filter → Item
Search–filter mismatch
Search results · Filter sidebar
3“handbag lv” appears as a text filter.
4Louis Vuitton is still unchecked.
Limited condition detail
Product page · Image gallery
6Seller photos provide limited close-up evidence.
Missing category feedback
Search overlay · Category tabs
5Switching tabs did not clearly signal updated results.
Key Findings
Purchase
- 1Unclear cart contents
- 2Incomplete cost information
Uncertainty about the order.
Product discovery
- 3–4Search–filter mismatch
- 5Missing category feedback
- 6Limited condition detail
Uncertainty across search, filtering, and item evaluation.
RESEARCH PRIORITY
Our review covered discovery and purchase. Testing focused first on:
Finding a suitable item and moving toward purchase.
Testing Hypothesis
RESEARCH HYPOTHESIS
Clearer navigation could make finding a specific item easier and more accurate.
We expected lower cognitive effort and greater confidence in continuing toward a purchase.
Moderated Study
Observe how shoppers find a match.
Understand the paths people take, the information they trust, and the moments that interrupt a purchase decision.
RECRUITMENT
Participant Profile
We recruited people with recent online secondhand shopping experience to compare the interfaces against familiar shopping habits.
18–40Target recruitment age
Past yearOnline secondhand purchase
10Moderated study
participants
participants
P1P2P3P4P5P6P7P8P9P10
Prior use of thredUP or Depop was not required.
Moderated comparative
usability testing.
The same shopping tasks on two existing websites.
+- 01
Think aloud
Hear what shoppers expect before they click.
- 02
Observe & record
Capture routes, hesitation, re-checks, and abandonment.
- 03
Follow up
Ask what made the item feel suitable—or uncertain.
Study Tasks
Find a secondhand jacket for an upcoming presentation, then repeat using a different discovery method.
TASK 01Preferred discovery method
Find a suitable jacket.
Black≤ $80Size MNear-new
Observe where shoppers begin and how they narrow the results.
TASK 02Different discovery method
Find an alternative.
Brown≤ $50Size MNear-new
Observe how shoppers change approach and compare options.
WHAT WE OBSERVED
Entry pointFilter useRepeated checksDecision confidence
My analysis process
From session observations to user patterns.
I used the Rainbow Sheet to compare behavior across participants, then Affinitization to organize related observations into themes.
- 01 / OBSERVE
Capture the behavior.
Where shoppers started, which controls they missed, and what they checked before deciding.
- 02 / RAINBOW SHEET
Compare across sessions.
Look for repeated behavior while keeping participant-level differences visible.
- 03 / AFFINITIZATION
Group and interpret.
Connect related observations into themes to discuss with the team and focus the next test.
AN EXAMPLE FROM THE REPORTED FINDINGS
Search-first did not mean friction-free.
The report identifies search as the starting point for seven participants and filter hesitation for nine. Together, these patterns directed our attention to what happened after entering a query.
View the reported user patterns ↗User Patterns
Moderated study · 10 participants · Existing websites
STARTING BEHAVIOR / SEARCH
70%7/10 participants
Started with search.
black jacket
ColorSizePriceCondition
9/10 hesitated with filters.
When applying filters on thredUP.
5/10 inspected condition in photos.
Photographs helped them judge product quality.
5/10 relied on images to decide.
Participant groups can overlap.
thredUP Findings
Three usability issues observed on the existing thredUP website.
Category scope was unclear
Shoppers expected men’s clothing in a catalog organized around Women, Girls, and Boys.
Business context: why Kids?
WHAT WE LEARNED ABOUT THE BUSINESS
During the project, we learned that thredUP had previously sold men’s clothing before narrowing its focus to women and children. We therefore framed Kids as a clearer signal of the available category, rather than a proposal to expand inventory.
The study does not establish whether participants’ expectations came from prior experience with the platform.
Size system mismatch
The task specified Size M; numeric labels required shoppers to interpret a different size format.
Filter discoverability
Shoppers scrolled through products before locating the size filter.
Study boundary: the tasks specified Size M, so the findings concern that conversion task. Inventory gaps also affected completion; they should be distinguished from interface friction.
Prioritize the core shopping flow
We prioritized the moments that interrupted the shopping task: finding filters and interpreting sizes.
WHERE WE FOCUSED
- 01
Search for an item
7/10 started with search
- 02
Find the filters
4/10 initially missed size
PROTOTYPE FOCUS - 03
Interpret the size
6/10 hesitated with size filters
PROTOTYPE FOCUS - 04
Evaluate the item
Condition detail remained a recommendation
Prioritize before redesigning.
Finding filters and interpreting sizes interrupted the task, so I focused the Maze prototype on those moments.
In our shared prioritization, I focused on where shoppers could no longer move forward confidently: locating the filters and interpreting sizes. I translated those priorities into the Maze prototype.
We explored the controls directly involved in the task. Richer product photography remained a recommendation, outside the isolated prototype evaluation.
Let behavior change the prototype.
Participants skipped the guide or clicked its examples, so we tested simpler choices and clearer guidance in the next round.
I built and iterated the Maze prototype. When participants skipped the guide or tried to click its size examples, I used those findings to revise the guidance and size-selection experience with the team.
The next round explored simpler size choices and clearer explanations. My focus shifted from providing more instructions to making the choice easier to understand.
Impact × Effort
Qualitative assessmentHigher impactLower impactLower effortHigher effort
01Category clarity
02Filter visibility
03Size clarity
04Condition detail
The matrix records the team’s qualitative assessment. Effort was a research-team estimate, not an engineering estimate.
06 / Unmoderated study
Iterative Prototype Testing
Two rounds of testing, with a revision between them.
- 01
Round 1
Test the initial prototype in Maze.
- 02
Identify issues
Review behavior and participant feedback.
- 03
Revise the prototype
Adjust the filters and size guidance.
- 04
Round 2
Revisit size selection and guidance.
07 / Unmoderated prototype test
Round 1
Explore filtering and size guidance in a shopping scenario.
6 participantsMaze · Unmoderated1 task · 2 follow-ups · 1 multiple choice
THE TASK
Find a jacket for a presentation.
ZaraSize MNearly New
Add a matching item to the cart.Stop before payment.
Filter use
Find and apply the controls.
Size understanding
Identify the right size.
Ease & confidence
Rate the task and share feedback.
Behavior changed what we tested next.
The guide was skipped.
Participants bypassed the introductory popup.
NEXT TEST
Explore size selection that does not depend on completing the guide.
Examples looked like controls.
Participants tried to select the size examples inside the explanation.
NEXT TEST
Clarify the guide and explore size information at the point of selection.
The task exposed extra choices.
The jacket task still presented size groups for other product categories.
NEXT TEST
Reduce irrelevant options and revisit how participants understand the size groups.
09 / Unmoderated prototype test
Round 2
Evaluate the revised shopping task with six new participants.
6 participantsMaze · Unmoderated1 shopping scenario · 2 task conditions
THE TASK
Find a jacket for a presentation.
ZaraSize MNearly New
Find a matching item and add it to the cart.Stop before payment.
Size selection explored in Round 2
Simplified size filter
Reduce irrelevant options and show size information within the filter.
Guidance with the previous filter
Explain size conversions before using the previous filter system.
What the testing told us
Make the guide clear.
Onboarding was not used by everyone. Participants who used it preferred a clear explanation.
Simplify the size choice.
The study reports that fewer size options were easier to understand. This supports further exploration of simpler selection, without declaring a winning version.
Check discovery and use.
Across both rounds, 10/12 used filters and 9/12 found or applied the size filter. These observations do not isolate the effect of filter placement.
Study setup & evidence notes
- Participants
- 12 across two rounds: P1–P6 in Round 1; P7–P12 in Round 2. Round 2 used new participants.
- Round 2 tasks
- One shopping scenario was repeated under two conditions. The report calls simplified filters without onboarding “Task A,” and the previous filters with onboarding “Task B.”
- Task order
- The study plan assigned P7–P9 to A → B and P10–P12 to B → A. AB and BA identify task order, not a result or a winning design.
- Follow-up
- Round 2 included three follow-up questions and two multiple-choice questions. The displayed Maze question asks which task made selecting and confirming a size easier.
- Count check
- The report summarizes size-selection feedback as 5/6, but its two embedded charts show four and three responses. Without participant-level records to reconcile them, that fraction is excluded from the results above.
- Scope
- 10/12 and 9/12 combine both rounds. “Found or applied” is the original combined measure; it does not establish correct size selection. Changes to guidance and filters were not isolated.
10 / Research synthesis
From explaining more to testing understanding.
Skipped guidance and mistaken clicks shifted our attention to what shoppers could understand while making a selection.
The tests informed the next research questions and interface recommendations. They did not establish improvements in speed, accuracy, or conversion.
11 / Research implications
What we recommended exploring.
Make filters easier to find, clarify size choices, and communicate the available category scope.
01 / RECOMMENDATION
Navigation
Consolidate Girls and Boys into Kids to make the children’s category explicit within the platform’s focus on women and children.
BEFORE / CURRENT
Separate Girls and Boys entries
AFTER / RECOMMENDED
One Kids entry
Context: the project identified a shift away from men’s clothing. Kids clarifies the proposed category scope; its effect on mistaken navigation still needs validation.
02 / RECOMMENDATION
Filter Placement
We recommend placing the main filters above the product grid so shoppers can find them while browsing.
BEFORE / CURRENT
Sidebar filters
AFTER / RECOMMENDED
Horizontal filter bar
4/10 moderated participants initially missed the size filter. The prototype rounds did not isolate horizontal placement from other changes.
03 / RECOMMENDATION
Size Clarity
We recommend grouping Standard Size and Plus Size and showing letter and numeric sizes together.
BEFORE / CURRENT
Mixed size options
XLXXL0X1X2X3X4X5X0002468101214161820222426283032One Size
BOTTOMS
XXSXSSmMedLgXLXXL0X1X2X3X4X5X0002468101214161820222426283032One Size
Interface detail · redrawn for readability
AFTER / RECOMMENDED
Clear size groups and labels
STANDARD SIZE
XXS(00)XS(0–2)S(4–6)M(8–10)L(12–14)XL(16–18)XXL(20–22)
PLUS SIZE
0X(12–14)1X(16–18)2X(20–22)3X(24–26)4X(28–30)
Interface detail · redrawn for readability
6/10 moderated participants hesitated with size filters. The displayed size mappings are prototype examples; brand, category, and regional sizing rules require validation.
FURTHER RECOMMENDATION
Add close-up photography of fabric, wear, and measurements. This recommendation still needs a dedicated evaluation.
NEXT STEPS
Separate the changes. Test the complete task.
Evaluate filter placement, size grouping, and guidance separately. Record discovery, correct selection, task completion, time, and errors in a working flow, including recovery when no items match.
12 / What I learned
Reflection
Start with the core task.
Use frequency, task impact, and effort to choose where to start.
I learned to resist redesigning everything at once. Moving from the broad heuristic review to a focused shopping task helped me prioritize the moments that blocked progress. I now use frequency, task impact, and effort to decide what deserves attention first.
Give behavior more weight.
Watch for pauses, retries, and missed controls—not just what people say.
Moderating sessions taught me to give behavior more weight: where shoppers paused, retried, or overlooked a control. Across the study, skipped guidance and attempts to click size examples showed why explanations alone were not enough. Comments helped explain the behavior.
Ground each change in research.
Connect every recommendation to an observation, then challenge it in the next test.
Building and revising the Maze prototype helped me connect evaluation with design decisions. I learned to link each recommendation to a specific observation, then use the next test to challenge it. Small prototype studies can guide a direction; stronger performance claims need clearer measures.