Critique Framework

A critique framework is a scripted operating system for design reviews that replaces chaotic opinion dumps with a repeatable sequence that actually improves the work. It rests on the three layer model every useful comment must follow. Start with raw observation by naming exactly what appears on screen such as the primary CTA sitting below the fold on a 1440 pixel desktop viewport or the error state red clashing with the success banner on mobile. Move to interpretation by linking that detail to consequences for users the brand or the system like mobile users scrolling past the CTA before understanding the page action based on February 2024 Hotjar data from a Shopify checkout flow. End with suggestion by offering optional moves such as testing a sticky header or swapping to a neutral orange but only after the first two layers are stated. Three frameworks map to specific project moments. Goal Rationale Feedback keeps reviews tight by making the designer state their goal in two sentences walk through their rationale in under three minutes and then restricting every reviewer comment to that exact goal which Vercel installed across squads in 2024 to cut review time in half. I Like I Wish What If pulled from IDEO workshops in the early 2010s works best early when the work stays exploratory because it demands positive observations first then wishes phrased as desires and speculative what if directions that prevent the room from killing momentum. The Decision Stack dominates late stage checks by requiring the designer to declare the exact decision needed list rejected options surface risks and offer a recommendation so the team tests the thinking rather than generates new ideas from scratch. Add six ironclad rules that the designer scopes the feedback in advance only one conversation runs at a time every comment must be specific no prescriptions land without a preceding observation everything gets captured in a shared doc and the meeting stays time boxed with a stated outcome at the end of each slot and the whole thing becomes reliable instead of exhausting. The facilitator rotates every sprint to stop any one senior from dominating and the designer learns to receive critique by listening fully before responding parking defensive reactions for 24 hours separating valid observations from bad suggestions and owning the final call because the framework informs but never commands.

This is not the senior designer monologue that still eats up calendars at most startups where one voice talks for twenty minutes about button radius while juniors nod politely and the work ships unchanged. It is not design by committee or a popularity contest where the loudest opinion wins. It is not personal taste wearing a fake mustache of principle such as hating rounded corners and calling it a usability crisis without tying it to any user test data from the last quarter at Intercom. A critique framework is not a presentation where the designer defends every pixel nor is it the optional meeting tacked on at the end when the file is 90 percent done and real changes would blow the deadline. If no one captures the layers decisions and follow ups in Notion or Coda the session evaporates and the next critique repeats the same ground. It is not a free for all that tolerates side conversations or general statements like this feels off. In 2026 remote async reality the framework is not a comment thread graveyard in Figma or a Loom video with 47 unmarked timestamps. AI generated variants do not get a pass just because they came from a machine. Without the script those outputs stay unaccountable and cross functional rooms full of engineers and PMs check out the moment designer jargon replaces plain observation.

Concrete examples show exactly how this lands in real teams. During Notion's 2024 AI command palette redesign the squad ran Goal Rationale Feedback on the mid fidelity mocks. The designer stated the goal was cutting average command selection from 8.2 seconds to under four based on the prior quarter usage logs. They explained the rationale for icon placement against the 2023 design system audit. A reviewer observed the palette collapsed on 13 inch MacBooks. The interpretation connected it to writers losing context during long editing sessions. The captured suggestion to add a pin toggle got built overnight and the next round of tests showed a 41 percent speed gain. At Stripe in 2023 the Radar dashboard team used Decision Stack for the final pre ship gate. The designer named the required decision on inline banners versus side panels for fraud alerts showed the three rejected options including a modal that had failed with 1200 merchants in 2022 tests and listed risks around alert fatigue drawn from their own metrics showing 29 percent drop off past three alerts. Reviewers probed the weighting and the notes logged live in FigJam produced a hybrid solution that cut false positive complaints 33 percent after launch. A healthcare startup in 2025 brought in a freelancer to run I Like I Wish What If on an early patient portal flow. Reviewers said they liked the card layout that matched patterns from the Mayo Clinic app they wished the appointment picker needed only two taps on Android and asked what if a calendar heatmap replaced the list entirely. The collaborative tone generated three new directions the original designer had missed and the portal shipped two weeks early with Net Promoter Scores 19 points higher than the previous vendor build.

Use critique frameworks at three fixed points per project when changing direction still costs nothing. Run I Like I Wish What If right after kickoff to confirm the team solves the right problem before anyone wastes weeks polishing the wrong one. Switch to Goal Rationale Feedback at mid fidelity once structure exists so structural flaws get fixed before high fidelity locks them in. Deploy Decision Stack pre ship when pixels near final form and the room must pressure test risks against real Amplitude or Mixpanel data. They become mandatory in remote first teams relying on Loom and Figma because async without the script turns into unread noise. AI accelerated loops need them even more since generation is instant but accountability still requires human eyes following the three layers. Rotate the facilitator so juniors build structural muscles and seniors learn to listen instead of perform. Never run them without hard 15 to 30 minute time boxes per piece without captured notes or without a stated outcome at slot end because the meeting then produces zero value. Skip frameworks on solo freelance gigs with no peers or during true production fires where you ship the fix first and run a post mortem later. Teams that let general comments slide or allow the designer to interrupt with defenses watch the best framework collapse into the same dreaded weekly meeting that improves nothing.

Critique frameworks replace loud opinions with a shared operating system so your team ships work that gets sharper instead of polishing the same mistakes faster.