Levi's
Fit + Style QUIZ
Background
PROJECT
I recruited and led a team of business leaders through an annual Levi Strauss & Co Hackathon, a weekend long event that brought cross-functional teams including marketing, design, development and data science together to ideate and build out innovative business ideas with high impact. At the end of the event, we presented and demoed our Interactive Marketing Quiz in front of eCommerce leaders.
I drew inspiration, research and learnings from the Micah Fit & Style Quiz.
EXECUTIVE SUMMARY
Build a digital, interactive and personalized fit & style quiz that helps Levi's consumers find head-to-toe products that align with their size, fit, and style preferences.
TEAM
As the key stakeholder and business lead of this project, I presented the idea to several employees at Levi Strauss & Co. and recruited a team to build the quiz:
-
Marketing & UX (myself): design strategy, business lead
-
Data Scientist: create algorithm logic
-
FED Partner #1: implement data logic into quiz
-
FED Partner #2: develop quiz front end
-
Visual Design: custom graphics, design consultant
Process
Research
Competitive Research, Existing Knowledge
Strategy
Features, Requirements, User Flows
design
UX Design, Visual Design, Custom Graphics
Dev + Finalize
Dev, Data, Prototype, Exec Presentation
Research
COMPETITIVE RESEARCH
Prose, Warby Parker and Stitch Fix all leverage easy-to-use quizzes that help shoppers find products that match their preferences on (3) factors: size, fit, and style. Of the online quiz market, few deliver personalized recommendations based on all three of these factors and across multiple clothing categories. This represents a golden frontier for Levi’s, in which shoppers may be more quickly connected with products that not only match their style preferences but also fit them well.
PERSONAS
Given the fast pace of the Hackathon, no documentation of personas were created, however we worked with merchants and our team’s data scientist to build (8) style profiles of consumers on the Levi’s website based on aggregated shopping data. We included factors such as seasonality and trendiness of the products they shopped, size and age variables, average order value, and the types of categories and fits they browsed.
existing knowledge
Given the short turnaround of the Hackathon, I built off research from the Micah Fit & Style Quiz to inform the Levi's Wardrobe Experience. The following are the "needs" cited by shoppers.
MAIN PROBLEMS
For this project, we wanted our tool to decrease the exit rates of consumers on site. We also wanted to reduce the amount of time it took for a consumer to add something to cart.
Strategy + ux
Features + requirements
To satisfy Levi’s needs, we’ll want to build a quiz that is:
-
Mobile friendly
-
Include questions for fit, size and style
-
Leverages existing in-house recommendations for Levi’s
-
Includes 2 user flows for both genders: men & women
-
Mirrors Levi’s branding
Based on the Micah Fit & Style Quiz, I determined the following design elements should be integrated into designs:
-
Image-based questions
-
Custom graphics to indicate shape or fit
-
Multiple-choice responses
-
Navigational Progress Bar
USER FLOWS
Consumer research has indicated that many men start shopping with fit in mind first, then size and then style. However, in a prior project I had found that consumers preferred to group the questions by product item type (pants, shirts, shoes), so I used this approach for the User Flow.
MARKETING
I switched on my marketing hat and collected all the relevant size, fit and style information needed for the quiz. This included a list of fits; each gender's style profile (and relevant products for each profile); and size ranges across bottoms, tops and accessories. This required collaboration with our merchants, and pulling information from our website and internal tools.
After collecting all the relevant information, I wrote sample questions for each of the quiz screens so they could be added into the designs.
Then, I found marketing assets that mapped back to our (8) Style Profiles and created a Shot List of the products they were wearing so that our Data Scientist could start building the product recommender system.
DESIGN
ux design
I designed men’s and women’s quiz experiences using the questions I had written, finding the right ways of displaying size bubbles, selections, and buttons.
PRE-HACKATHON
Wireframes + Mocks V1
Building off the original mocks I created from the Fit & Style Quiz, I adapted the quiz framework and used Levi's branding including font face. design elements like the "squiggle"
visual design
I passed off the wireframes to our Visual Design expert, who helped me refine the font weights and color treatments. With those wireframes, she started slicing assets for each quiz experience — men and women — based on the image recommendations and started designing custom graphics for our designs, including the fit drawings and our unique loading animation between each screen.
HACKATHON WEEKEND
We had 1.5 days to implement design and data. We went through 2 rounds of QA but prioritized
Development + finalization
Development + Data
After handing off the assets , our developers and data scientist worked to code both quizzes and connect the customized recommendation algorithm to our backend using JSON outputs. For the MVP, we used a manual recommendations approach, where each response produced pre-set recommendations however with more time and energy we’d have the ability to automate the recommendations system.
Presentation + Prototype
We presented our prototype with LS&Co eCommerce leaders. Check it Out!
Video: Men's Experience
Video: Women's Experience
closing thoughts
If we wanted to invest in building this quiz out further, the following would be my recommendations: |
-
Account for size & inventory data - not doable during Hackathon weekend, but would be valuable for productionalized product
-
Control for Large File Sizes - the recommendations data file is large for hybris; need a way to systematically condense the data
-
Create quiz flows for other size groups - this includes Big & Tall, Plus sizing
-
Testing of All Response variations - to ensure that all customers get quality recommendations, test all variations of test responses and develop rules that prevent users from not receiving recommendations data
-
Improve models for recommendations - Can test out training off other models, incorporating heuristics into filtering of final results
-
Segmentation of Style Profiles - As recommendations improve, will have more data around what expressed preferences differentiate consumers which can inform style options