Making ML Competitions More Approachable

Role: Lead Designer

Team: Product Manager, Engineer, Senior Designer

Launched April 2022  View live

The product

Kaggle is a platform that offers machine learning competitions across various categories of interest and skill level. Users build their ML skills by competing in competitions and learning from the community.

Design System Impact

The new component I designed became a useful addition to the design system. Teammates have reused the component as a guidance tool on Kaggle’s homepage and as a work organization tool on users’ profile page.

Proposing a Solution

Identifying this as an opportunity to improve the UX for new users, I decided to propose a new feature. I wrote a Project Requirements Document that highlighted the user need, my proposed solution, and the project’s alignment with business goals, and presented it to senior stakeholders. They agreed that this was a problem worth solving, and approved the project.

Identifying an Opportunity


Noticing a Problem

This project sprang from a heuristic evaluation of Kaggle that I conducted. While exploring the site, I noticed that the Competitions landing page divided content into unclear categories. As a newcommer, I had no idea what a “Playground” or “Community” competition meant.

Validating the User Need

I suspected that other new users were similarly confused. Upon investigation, I discovered several new users in the Kaggle forums seeking clarification on our categories:

“What is Tabular Playground? and is this for beginners or intermediate/advanced?” - User in Kaggle forum

I also found past user research stating that Kaggle-specific jargon was a big barrier to new user engagement. New users get intimidated and confused by the internal lingo, which makes it harder for them to use the platform.

Ideation


Brainstorming

To brainstorm ideas, I did a Crazy Eights rapid sketching exercise while keeping 3 How Might We questions in mind:

  • HMW make categories clear and helpful?

  • HMW introduce new users to different types of competitions?

  • HMW direct users to the right type of competition for them?

Early Explorations: Going Broad

I went very broad in my early wireframing, exploring some light solutions that required minimum UI changes to more radical solutions that necessitated restructuring our categories or the entire landing page.

Option1: Shelves

Restructures landing page to give each category its own section with a description + content preview

Feedback: Favorite amongst design team for its glanceability, easiness to browse, and UX consistency with other pages

Option 2: Mini Info Cards

Expands current chips to more informative filter cards

Feedback: Favorite amongst stakeholders for its simplicity, scalability, & encouragement to explore

Winner!

Option 3: “Learn More” button

Opens up a dialogue explaining our competition categories

Feedback: Least popular; too heavy

Guiding Questions

  • Hierarchy: How much attention do we want to draw to this section?

  • Visual: How might we incorporate visual intrigue? What are the pros and cons of adding color?

Narrowing Down Options

To decide which design approaches to further explore, I evaluated each approach against design criteria that included scalability, discoverability, and my original HMW goals. The solution options that met the most criteria were:

  • Info Cards

  • Shelves

  • Learn more

Team Feedback

Next Step: Present wireframes of these three options to the design team and to project stakeholders for feedback.

Choosing Design Direction

Post-Feedback Iterations

Adding the New Component to our Design System: Defining component specs & behaviors

The Problem

Platform-specific language confuses new users and creates a barrier to engagement. 

With competitions organized into unclear categories, new users struggle to find a suitable competition to join.

The Goal

Help users understand each category's meaning so that they can quickly find a competition that suits them.

Component Specs & Handoff

Scope creep example: What if we completely recategorized our competitions into three skill levels - Beginner, Intermediate, and Advanced?



With my design approach selected, I could now zoom into the details and experiment with hierarchy, iconography, brand colors, UX writing, and more. I needed to figure out how to make this design useful to both novel and veteran users.

Questions

  • Interaction: Should this section appear for all users, or just newbies? Should it be dismissible?

  • Interaction: What happens when the user clicks on the card? What happens if multiple filters are applied?

Scope Creep: Restructuring Categories

The original goal of this project was simple: explain what each category meant. This goal, however, was a very zoomed-in solution to the user’s broader need: to easily find the right competition for them. Was doubling down on our current categories the right solution, or were we solving the wrong problem?

I started exploring more creative ways to guide users to suitable competitions, including restructuring of our categories to be skill-level based instead of topic-based. While this solution had unique strengths, it required more engineering and product restructuring than planned. After realigning with stakeholders on goals and scope, I eliminated this option as out of scope.


Post-Implementation Adjustments: Notes to engineer

Impact


User Impact: More Exploration, Less Intimidation

With more informative and inviting competition categories, exploration of our competition offerings through these filter chips has increased by 9%.

Previous
Previous

Google: File organization system for data scientists

Next
Next

UX Research & Strategy for my Inclusive Travel Startup