A frontend build can look correct and still feel wrong.
This usually happens when the team receives finished designs but not finished interaction decisions.
The design handoff might include:
- screens
- components
- spacing rules
- design tokens
- responsive layouts
- prototype links
That is useful, but it is not always enough to build a product that feels clear.
The frontend still needs answers to questions like:
- What happens before data loads?
- What happens if data is empty?
- What should the user see after an error?
- Which state is optimistic and which state waits for confirmation?
- When should the route change?
- What should be handled by the component, the server action, the API, or the CMS?
- What copy explains uncertainty?
I think of this as interaction architecture.
It sits between UX and implementation. It defines the behavior that makes the interface usable after the visual design is done.
Example: a multi-step lead form
A Figma file might show each step clearly. But the real frontend still needs to decide:
- Can the user go back without losing data?
- Is validation immediate or after submit?
- Are optional fields visually different?
- What happens after refresh?
- Does progress save locally or server-side?
- What does the confirmation screen promise?
If those answers are missing, React components become the place where product strategy gets improvised.
That is risky because engineering decisions start shaping UX without enough context.
A simple checklist before implementation
For every important flow:
1. List user decisions.
2. List system responses.
3. Define loading, empty, error, partial, and success states.
4. Decide where state lives.
5. Write uncertainty copy before styling deeply.
6. Prototype the behavior, not only the screen.
This is also why design and development should not be treated as unrelated phases. The best web products are planned as one system: intent, interface, state, and implementation.
Reference from our development work at MDX:
https://mdx.so/development
Disclosure: drafted with AI assistance and reviewed before publishing.
Top comments (0)