Design reviews should critique decisions, not pixels
The most expensive mistakes in a design review are the ones everyone is too busy adjusting margins to notice.
Most design reviews fail before anyone opens their mouth. They fail because the artifact on the screen invites the wrong conversation. A pixel-perfect mockup begs for pixel-level feedback, and a room full of capable people will happily spend forty minutes relocating a button by four pixels while the actual decision — should this screen exist at all — sails past untouched.
The job of a design review is not to make the work prettier. It is to pressure-test the decisions the work encodes. Every interface is a stack of choices: what to show, what to hide, what to make easy, what to make deliberately hard, who this is for and who it quietly excludes. Those choices are where the money and the risk live. The margins are downstream of all of it.
The reviews that cost the most
The expensive mistakes are never the ones that look wrong. They look fine. A signup flow that converts beautifully and trains every user to ignore the one permission screen that matters. A dashboard that's a joy to read and answers a question nobody actually asks before they act. A settings page so clean it hides the fact that the default leaks data. None of these will get flagged by someone squinting at letter-spacing, because they aren't visual problems. They're decisions wearing a nice coat.
I've sat in reviews where the team adjusted the same hero image three times and never once asked whether the page needed a hero. The polish was real. The work was confident. It was also solving a problem we'd never agreed was the right one. That's the failure mode that should scare you — not ugly work, but plausible work pointed in the wrong direction.
Polish is the easiest thing to see and the cheapest thing to fix. Spend the review on what's neither.
Review the decision, then the execution
The fix is structural, not a matter of willpower. You change what the review is allowed to be about, and you change it in that order: decisions first, execution second, and never the two interleaved.
Start every review with the designer stating the decisions, not demoing the screens. What problem is this solving, for whom, and what did you choose to make easy at the expense of what else? What did you cut, and why? Now the room is evaluating a position, and a position can be argued with. A screenshot can only be reacted to.
- →Ask "what should this make easy?" before "does this look right?"
- →Treat every default as a decision, because to the user it is one.
- →When someone reaches for a pixel, ask what decision the pixel is serving. Usually there's a real one hiding underneath, and that's the thing to discuss.
Only once the decisions hold up does execution feedback earn its place. At that point margins and type and spacing matter enormously — craft is how a good decision survives contact with a real user. But critiquing the execution of a decision you haven't validated is how teams build beautiful versions of the wrong thing, fast.
Make the structure do the work
You cannot win this with discipline alone, because the pull toward pixels is gravitational — it's the most concrete, most agreeable thing in the room. So build the review to resist it.
Put the problem statement at the top of every review doc, in the designer's own words, and read it aloud before the screens appear. Present flows and decisions before high-fidelity comps; a grayscale wireframe gets you a conversation about structure because there's nothing else to talk about. Separate the two passes explicitly — "we're reviewing the approach today, fidelity next week" — so nobody feels they have to choose between the two and defaults to the easy one.
And write down the decisions you reach, not just the changes you request. "Move the CTA up" is a fix. "We're optimizing this screen for returning users, so the primary action is renewal, not discovery" is a decision — and it's the thing the next three reviews should be measured against. Teams that record fixes relitigate the same arguments monthly. Teams that record decisions compound.
The best reviewers I've worked with barely mention the visuals until the end. They ask what you're trying to do, who loses when you do it, and what you'd have to believe for this to be right. By the time they get to spacing, the hard part is already settled — and the spacing, freed from carrying the weight of an unexamined decision, gets fixed in two minutes.
Pixels are not the enemy. They're just the cheapest thing in the room, and cheap things crowd out expensive ones when you let them. Run the review so the costly questions go first, while there's still time and goodwill to answer them. The margins will wait. The decision won't.
