How I work
There’s a nice diagram: Empathize → Define → Ideate → Prototype → Test. Arrows, circles, everything in its place. In five-plus years, I’ve never once seen a project actually go that way.
TL;DR — I figure out the context → name the problem → agree on what success looks like → come up with three different paths → pick one and test it cheaply → get it to launch with the team → look at what happened and draw a conclusion. It works for a startup from scratch and for a button in settings.
Before I open Figma
I’m rarely brought into a project with a clean slate. So the first thing I do is map the terrain: what’s already been tried and quietly buried, who actually makes the call (not on the org chart — in practice), what’s really hurting the business behind the ask, and what nobody is going to rewrite under any circumstances.
Then I go find the actual problem. On a new product I have no data, so I talk to people: six to eight interviews is usually enough for patterns to repeat. I never ask “would you use a feature like this” — people aren’t lying, they’re just bad at predicting their own behavior.
From sketch to shipped
I think in flows, not screens: where the person is coming from, what they already know, what happens next. Screens are frames. The product is the film. And I work the edge cases early rather than at the end — empty states, real error messages, slow connections, 500 rows instead of three, someone without permissions, a name three times longer than the mockup expects. A product feels cheap not because the home screen is badly drawn, but because nobody thought about the second Tuesday of using it.
Then I validate in order of increasing cost: show the engineers (ten minutes tells me what’s two sprints and what already exists as a component), hallway test, then real users, then production. The rule I learned the expensive way: I’m not testing my mockup, I’m testing my hypothesis. Otherwise I catch myself hunting for confirmation and telling people where to tap. Staying quiet is physically hard. It’s a skill, and it trains.
UI comes last, once the logic holds up — design system components doing the work, not inspiration. There are still lots of iterations.
