I used to think finishing was just more of the same work, done slower. Turns out it’s a different job entirely, and I only really understood that after watching two very good designers fail at it in opposite directions.

The generator and the finisher

One of them could produce three credible directions before lunch. Fast, loose, confident — the kind of designer a project needs in week one. Put the same person in week six, polishing a shipped flow’s edge cases, and the work stalled. Not because they got worse. Because the skill that made week one fast — generating options, moving on before doubt set in — was actively working against week six, where the job was to sit with one option and find everything wrong with it.

The other designer was the reverse. Slow to start, uncomfortable committing to a direction before it felt right — and then, once something existed, relentless. Every spacing value questioned, every transition timed against three alternatives, every edge case chased down. Early on I read this as indecisiveness. It wasn’t. It was someone whose actual skill only turns on once there’s something concrete to interrogate.

What the first ninety percent actually rewards

Getting from nothing to mostly-there rewards speed of judgment. You’re constantly deciding “good enough, move on” across dozens of small choices, because stopping to perfect any one of them means the other ninety don’t get made at all. The skill is knowing which ten things matter enough to slow down for, and moving fast past the other ninety.

This produces real value. A rough version of the whole thing, end to end, tells you more about whether the concept works than a polished version of one-tenth of it ever could. I don’t think this phase is optional or lesser. It’s just a specific skill, and it’s not the same one as what comes next.

What the last ten percent actually rewards

Finishing rewards the opposite instinct: the willingness to stay on one thing past the point where it feels done, because “feels done” and “is done” are different claims. The corner radius that’s technically fine. The loading state nobody designed because it only shows up for 400 milliseconds. The focus ring that works for a mouse and disappears for a keyboard. None of this is generative work. It’s closer to proofreading — except the errors don’t announce themselves, and most people stop looking before they find them.

The uncomfortable part: this phase produces almost nothing that shows up in a portfolio. Nobody screenshots a corrected line-height. The work is invisible by design — you’re not adding anything, you’re removing the hundred small reasons an interface feels slightly off in ways nobody can name.

Staffing for both

The mistake I made for years was assuming a good designer was good at both, and that the same person should carry a project start to finish. Some can. Most, in my experience, are meaningfully better at one than the other — and a team that quietly lets its best generators also do the finishing, or its best finishers also do the initial exploration, is leaving real speed on the table in both directions.

I don’t have a clean system for this yet. What I do now is ask, out loud, in week one: who on this team actually enjoys week six? The honest answer is rarely everyone, and it’s almost never the person you’d guess from their week-one output.

Rhys O’Sullivan is a product designer at Fieldstone, previously Loop. This is one in an occasional series of notes on research, systems, and the process of making things.