← Atlas · Topics: cta, ux-patterns, landing-pages
Good UI
- Site
- goodui.org
- Type
- pattern library
- Verdict
- Useful
- Licence
- Conditional licence, proprietary paid
proprietary-paidPattern names, test write-ups and the original 75 ideas are free to read. The actual effect sizes are members-only: Solo $60/month billed yearly ($720, 1 user), Team $120/month billed yearly ($1,440, 5 users), and an Expert-Guided plan at $1,950/month with a monthly test review. The site offers a 30-day refund if a pattern doesn't work for you. The Terms forbid copying, reproducing or reselling any part of the service without written permission, and each plan licenses the data to a set number of users. - Agent channels
- none
- Pricing
- freemium
- Reviewed
What it is
GoodUI is run by Jakub Linowski, a Toronto-based experiment designer. It collects results from A/B tests that companies and practitioners share, and groups them into repeatable UI patterns. The patterns page counts 141 patterns from 642 tests: 168 winners, 43 losers and the rest not statistically significant. According to the pricing page, the tests cover about 149 million visitors. There is also a "Leaks" section on experiments spotted on sites such as Etsy, Amazon and Booking.com (112 so far), plus the site's original list of 75 UI ideas written before the testing began.
When to open it
- When you are choosing what to test on a landing page, product page, signup or checkout and want to start with ideas that have a track record.
- When someone suggests a change (sticky CTA, countdown timer, fewer form fields) and you want to know whether it has won or lost elsewhere.
- When you want a free list of conversion ideas to start a design review.
Most useful
- Browse by screen, metric and tag: home and landing, product, listing, checkout, pricing, cart, signup, thank-you; leads, signups or sales; forms, headlines, pricing, popups, buttons, social proof.
- Pattern pages: each shows how many times the pattern was tested and on which screens and devices, with A/B screenshots and a short note on each test.
- Statistical power is shown for every test, and the homepage can sort tests to only those with 80% power at a 5% minimum detectable effect. That makes it easy to skip weak tests.
- Losing tests are published too, so you can see which popular ideas tend to fail.
- The free 75 ideas: short "try X instead of Y" pairs such as a one-column layout, repeating the CTA on long pages, or social proof instead of talking about yourself.
Using it with agents
There is no llms.txt (404), API, MCP or export. Pages are server-rendered, so an agent can read the free pattern names and descriptions, but the win rates show as "X.X%" unless you are signed in. Don't paste members-only numbers into prompts or shared files, because the licence is per user. A safe approach is to name the patterns in your brief ("use pattern #41, sticky call to action, on mobile") and let the agent build them.
Watch out for
- The headline numbers only appear after you pay. For free readers, a pattern page shows what was tested but not whether it helped.
- A win on one site is not a promise for yours. The site's own FAQ says the results are about probability, not guarantees.
- Most tests come from ecommerce, lead generation and SaaS funnels, so the patterns lean heavily toward conversion.
- The homepage leads with a newsletter form and a strong sales pitch. You can ignore both and still browse.
Reusable ideas
- Show the sample size and statistical power next to every claim that a design "works".
- Keep losing and flat results next to winners so a pattern's track record is honest.
- Give every pattern a stable number and name so teams can refer to it in briefs and tickets.
- Write design ideas as "try X instead of Y" so each one is a testable swap.