All posts
AIProductResearch

From What Users Say to What You Ship: User Research as the Source of Truth for Your Design System

How user research has an impact on Design.md

Insha Naseem
Insha Naseem
Author
May 25, 2026
4 views 4 min read

Last week, we covered how to build a design system foundation - tokens, typography, spacing, and component architecture. But here's the question that keeps getting skipped: where do those decisions actually come from?


The Missing Link Between Research and Your Design System

A design system without research is just an opinion at scale.


Every token you define, every component state you document, every spacing rule you enforce - it either maps to a real user behavior, or it doesn't. And when it doesn't, your system quietly accumulates debt: components nobody uses, patterns that confuse, states that were never the right states.


User research is not a phase that happens before design. It's the continuous input layer that keeps your system honest.



What Research Actually Tells Your Design System

1. Tokens Come From Mental Models, Not Mood Boards

When you define color.feedback.error or spacing.component.gap, you're making a claim about how users perceive and process information.


Research validates that claim.


According to NN/G, user interviews are most powerful in the discovery phase - revealing who your users are, what their experiences are like, and what they need.


That's token-level signal. If users consistently misread a warning state as an error, your color.feedback.warning token is doing the wrong job - no matter how elegant the hex value.


Research input → Token decision:


What Users Tell You

What Changes in the System

"I didn't know it was clickable"

Review color.interactive.default contrast ratio

"The spacing felt cluttered on mobile"

Audit spacing.layout.dense against real viewport behavior

"I ignored that - I thought it was a label"

Reconsider typography.label vs typography.caption distinction



2. Component States Are Defined by Real Edge Cases

Last week's design.md documented component states: default, hover, active, disabled, loading, error.


But here's what research adds: the states users actually encounter vs. the states we assume they encounter are rarely the same.


NN/G notes that the distinction between "what people say" and "what people do" is one of the most critical tensions in research - and very often, the two are quite different.


A loading state that lasts 200ms in your test environment might last 4 seconds on a user's actual network. That changes everything about how your component should behave. Research surfaces those real-world conditions that documentation alone never will.


Before adding a component state → ask: Has a user ever encountered this? What did they do?



3. Patterns Are Validated, Not Assumed

Your design system documents patterns - form layouts, navigation structures, empty states, error flows. Research tells you if those patterns actually reduce friction or just look clean in Figma.


NN/G found that open-ended questions in interviews result in deeper insights - the kind that closed questions and analytics dashboards simply can't surface.


That depth is what separates a pattern that converts from a pattern that gets abandoned. Analytics tells you that users dropped off. Interviews tell you why - and that why is what you encode into your system.



The Prerequisites: What You Need Before Research Informs the System

Getting research to actually feed your design system requires more than good intentions. It requires a process.


Step 1 - Define the research question tied to a system decision. Don't do generalist research. Ask: "We're defining our notification component. What do users actually need to know, and how urgently?" That's a system-level question.


Step 2 - Recruit for behavior, not demographics. As NN/G's Director of Research Maria Rosala notes, targeting based on what people do - tools they use, tasks they perform - produces far more relevant signal than demographic filters alone.


Step 3 - Use AI to prepare, not to replace. Let AI draft your screener, recruitment email, and interview guide. Use that time to think deeply about follow-up questions. AI handles scaffolding. You handle judgment.


Step 4 - Map findings directly to system components. After every research session, run a 30-minute synthesis exercise with one question: "Which component, token, or pattern does this finding challenge or validate?"



How It Feeds Back Into design.md

Think of your design system as a living document with two update triggers:


Trigger

Source

Update Type

New product feature

Product roadmap

Additive - new component or variant

User research finding

Interview synthesis

Corrective - token, state, or pattern revision


The first is what most teams do. The second is what separates systems that scale from systems that slowly become irrelevant.


Your design.md isn't finished when it ships - it's a hypothesis. Research is what tests it.



The Design System Impact, Summarised

User Research Finding


        ↓


Mapped to Token / Component / Pattern


        ↓


Validated Against WCAG + Real-World Behavior


        ↓


Documented in design.md with Rationale


        ↓


Shipped with Confidence, Not Assumption



What's Next

Next week: How to run a design system audit - using your research findings as the lens to identify what's working, what's stale, and what's actively harming users.


In the meantime, pull up your design.md. Pick one component. Ask: "Do we actually know why this state exists?"


If the answer is no - that's your first research question.




References: Nielsen Norman Group - User Interviews 101, Accelerating Research with AI, When to Use Which UX Research Methods


ConveGenius Daily Signals

Receive the next signal

Get future product, design, AI, engineering, and team signals directly in your inbox. Only published signals. No spam.

Unsubscribe anytime · No tracking pixels

Reactions
Sign in to react

Discussion

0 comments
Login to comment. Use Google to join the discussion.
Login to comment
No comments yet. Start the signal.
More from Insha Naseem