Design System

Creating and deploying a web-first design system tailored to WCAG with 400+ components

roles

Design Systems

tools

Figma

Claude

team

2 Designers

3 Software Engineers

overview

Design and 0 -> 1 systems thinking

Forkast is a venture capital-backed B2B startup based in the restaurant competition and market analysis industry. Currently, it's used by hundreds of chains, including Michelin-starred locations like Benares and fast food stops, including Red Robin.

As a product design intern at Forkast, I co-led and executed two major initiatives: 1) redesigning the dashboard system used for analysis 2) building a design system from the ground up (shown here).

Their users were scaling up rapidly, but the designs couldn't keep up with the data.

A primary issue communicated by the team's primary engineer was that the code and their consumer base had scaled to 500+ restaurant chains (or thousands of restaurants) in a short time.

They had reliable data, but when they implemented new features or tried to adjust existing ones, establishing branding and consistency became an issue.

After I audited their internal dashboards, I also noted issues with the design of components through the lens of how intuitive they were to read.

As Digital Spaces Increase, Meaningful Connection Decreases. Or so it feels...

Social isolation in American adults rose from 11-20% to over 20% in 2024. XHacks proposed: how might we reconnect in the digital age? How can we devise an original solution combatting interpersonal disconnect over time?

I created a design system based in Figma, made for scalability and the Forkast context.

I advocated for a consistent benchmark: a Figma design system that could be linked with Claude Code.

I created the overarching structure of the design system, focusing on scalability as I anticipated that the data they would show would continue increasing, organizing how libraries referenced other libraries, what belonged in collections, and established consistent component naming conventions.

For the components themselves, I and one other designer created them from scratch, balancing interaction states (like hover, disabled, focus, etc) with some WCAG compliance I pitched (color contrast, text sizing).

Some of our design system is below [most hidden for NDA]:

As Digital Spaces Increase, Meaningful Connection Decreases. Or so it feels...

Social isolation in American adults rose from 11-20% to over 20% in 2024. XHacks proposed: how might we reconnect in the digital age? How can we devise an original solution combatting interpersonal disconnect over time?

Before Creating...

How did I get there?

Goal Alignment

Forkast's Dashboard and Quantity-focus

I spoke with Forkast's founding team to understand their working goals to then define constraints for myself before ever touching Figma.

1:. Forkast's platform is and will be data-heavy due to business constraints

_

Large restaurant chains have thousands of restaurants. [Their previous dashboards, below, are blurred for NDA]

Consequently, Forkast's design system would need to be flexible with high information density. I researched industry standards for design systems, as being data heavy did = different conventions for features like padding and font size.

2:. Forkast's scaling rapidly as a business. What does this mean for design?

+

Issue Analysis

Auditing Current Designs

I used an HTML to Figma plugin to pull current components, investigating potential pain points with their current Claude Design-generated system (besides the known uncertainty about consistency).

a.

Breaks in convention: Many components didn't follow standard design conventions, especially for accessibility guidelines.

b.

Molecular-level issues: Most were data-driven and not designed for readability, so the dashboards felt like cognitive overload, rather than actionable insight.

Part of the restructuring of the overall dashboard occurred at this foundational level, in making these molecules easy to consume.

Design System Structure

Design System Structure

Following industry standards, the global library will remain mostly untouched, while the semantic library is pulled into working design documents.

AI and Iteration

Efficiency with Claude

a.

The Context: I was exploring overall library structure + building components at the same time to understand how the two needed to interact for max efficiency.

b.

The Issue: Manually relinking colors to components for tests became a bottleneck: in Figma, turning a collection into a library and pushing it into a separate file = broken reference links.

c.

Explorations: To relink variables with components, I first explored several plugins (which didn't work). Relinking by hand would be too slow for steady experimentation and iteration…

d.

AI: I tested out using Claude to automate the relinking while I iterated through structural shifts.

e.

Result: Claude could infer which variables to use reliably, so I adapted it into my workflow. This in turn freed up time for me to attack other portions of the design system that required more thought + human decision-making.

Reflection

Cohesion + Continuity

Designing for cohesion across vastly different elements was an intriguing challenge, as I typically work with existing design systems.

This system is always evolving— as the team expands and continues to design, I expect to continue iterating! Fixing structural errors and continuing to adapt for new features are our likely next steps, and I'm excited to see how we evolve.

designed by joana liu :D