When I worked on the design system at Inbound, the sign it was working was a boring one.
Once the system covered the common cases, most screens stopped needing design decisions at all. The parts existed. The spacing was decided. They just needed assembly.
Designers can hear that as a threat. I think it's the best thing a system can do for them.
Count the parts
I wanted to know how true this is, so I looked at the 55 blocks in HextaUI Pro. They're complete pieces of product UI: a cart, an order page, a settings screen, a chat thread. The median block imports 10 components. Here's what shows up most.
A very different set of screens keeps reaching for the same handful of parts. An e-commerce block and everything else both land on a median of 10.
So at the level of parts, a new screen is mostly a question you've already answered.
Where the decisions should go
A new screen does still need thinking. The thinking just belongs somewhere else.
Without a system, each of those dashed boxes gets re-decided in a small way every time. How far is the title from the filters? What does this table say when it's empty? Each answer takes a minute, and nobody thinks of it as a real decision. By the end you've spent your attention on padding, and the part that matters got whatever was left.
A test for your own system
Take the last screen your team built and list every decision made while building it. Then sort them.
- What does this page help someone do?
- What matters most here?
- What happens if this action fails?
- Which grey?
- How much padding?
- What does the empty state look like?
- Which button style?
- How big is the title?
The fair objection
The obvious worry is that everything starts to look the same. It can.
But I've noticed that sameness tends to come from a different place: copying someone else's system without making any decisions of your own. Deciding once, carefully, is the opposite of that. Your type, your spacing and your colours can be as strange as you like. You just shouldn't have to pick them again on Tuesday.
A system also has limits. At Inbound, the dashboard and the marketing pages pulled in opposite directions, one dense and one spacious, and both had to sit on the same shared layer. Force everything above that layer to match and you get the worst of both.
What it changes
The goal I had at Inbound was that two people building two different screens would arrive at interfaces that look like the same product.
Once the basics stop being up for debate, the conversation can move to what the screen should do.
That's where a designer is worth paying for. Assembly can look after the rest.