Web Usability

Web Usability — Agent Skill (Don't Make Me Think)

Rules from Don't Make Me Think, Revisited (Steve Krug, 2014). Users scan and satisfice — design for that, not for careful reading.

Pair with web-api-design for backend contracts and clean-code for implementation.

When to apply

• Designing or reviewing web/mobile pages, flows, navigation, labels, forms
• Writing UI copy, error messages, onboarding
• Planning lightweight usability tests
• User says: usability, UX, Krug, don't make me think, make this obvious

Core laws

1. Don't make me think — every page/question should be self-evident at a glance.
2. Users scan, they don't read — design billboards, not essays.
3. Users satisfice — first reasonable option wins; make the right choice obvious.
4. Users muddle through — they use what works, not what's optimal; don't rely on instructions.
5. Omit needless words — half the words on most pages can go.
6. Test with real users — early, often, cheap; fixes beat debates.

Agent workflow

1. GLANCE TEST  — can user tell what this is and what to do in 5 seconds?
2. SCAN DESIGN  — hierarchy, headings, bold keywords, short chunks
3. CHOICES      — mindless, clearly labeled, no clever names
4. NAVIGATION   — persistent, breadcrumbs, you-are-here, search when needed
5. COPY         — cut words; kill happy talk and instructions
6. TEST PLAN    — 3 users, monthly, fix top issues

How users actually behave (Ch 2)

• Scan pages for words/phrases that match their task
• Click the first thing that looks close enough (satisficing)
• Back button is the most-used control — support undo mentally
• Never assume users read instructions, compare all options, or choose optimally

Design implication: make the right thing look clickable and obvious.

Billboard design (Ch 3)

• Clear visual hierarchy — what's most important looks most important
• Use conventions — logo top-left, nav top, search box where expected
• Break pages into clear areas — user decides where to look first
• Make clickable things obvious — buttons look like buttons; links look like links
• Format text for scanning — short paragraphs, bullets, bold key terms
• Minimize noise — every extra element competes for attention

5-second test: Show page 5 seconds — can they say what it's for?

Mindless choices (Ch 4)

• Every click should not require thought
• Category names must be mutually exclusive and unambiguous
• Avoid cute/punny labels — use what users would say
• Trunk test: On any random page, user should instantly know: site, page, major sections, local nav, utilities, "you are here"

Omit needless words (Ch 5)

Remove:

• Happy talk — "Welcome! We're glad you're here…"
• Redundant instructions — if UI is self-explanatory
• Half the words — then half again

Keep: words that reduce ambiguity or answer a question the user has.

Navigation (Ch 6)

• Persistent navigation on every page (except focused flows like checkout)
• Show location — breadcrumbs, highlighted section, page name
• Home link on logo; Search when catalog/content is large
• Utilities (login, cart, help) separated from primary nav
• Site ID + tagline on every page unless user is deep in a known section

Home page / landing (Ch 7)

Answer immediately:

1. What is this?
2. What can I do here?
3. What's new/important?

Don't cram everything — prioritize what most users need. Tagline = 6–8 words explaining what you do.

Mobile (Ch 10)

• Tradeoffs are real — less screen = harder; don't hide essentials
• Smart defaults, sensible autofill, forgiving inputs
• Learnable gestures only when payoff is clear
• Desktop patterns don't always translate — test on device

Courtesy & accessibility (Ch 11–12)

• Be a mensch — helpful error messages, don't punish users for mistakes
• Accessibility helps everyone — alt text, contrast, keyboard nav, form labels
• Fix obvious a11y issues in every review (headings, focus, link text)

Usability testing (Ch 9)

Do this:

• 3 users per round is enough to find major issues
• One morning a month — not a big lab study
• Give tasks; stay quiet; note where they hesitate or go wrong
• Fix top serious problems first; debaters lose to testers

Don't: wait for perfect prototype; argue instead of test.

Smells to flag

Review output format

## 5-second test
[Can user tell what this is and what to do?]

## Must fix
- [Blocks task completion or causes confusion]

## Scan & hierarchy
- [Headings, noise, clickable affordances]

## Copy
- [Words to cut or clarify]

## Navigation
- [Location, trunk test failures]

## Quick test suggestion
[1–3 tasks for next usability round]

Source

Steve Krug, Don't Make Me Think, Revisited (3rd ed., 2014).