Most Expensive Design Mistakes (Ever) and how to avoid them - Clarissa Rodrigues

Main ideas, concepts, and lessons

• Expensive design mistakes happen when teams don’t understand “why” a design fails

  • Clarissa’s career story illustrates a common evolution: early on, people believed they “knew what customers want” without research, resulting in products/UI that were later not used as expected.
  • The corrective mindset: focus on why usage drops (e.g., users don’t click the button you expect) and connect that to measurable evidence and UX research.

• Use evidence, not guesses—especially at scale (Uber context)

  • Uber’s advantage: large-scale behavioral data (usage metrics, click/engagement patterns, etc.).
  • This enables “reverse engineering” of failures: determine where users get stuck, what they ignore, and what information patterns work.

• People mostly scan, not read

  • Users:
    • scan in different directions/patterns
    • prioritize bold/important information first
  • Designers should support scanning with:
    • strong visual hierarchy
    • good core navigation
    • readability improvements like size, contrast, proximity, alignment

• Design clarity beats cleverness

  • “Be clear, not clever.”
  • Avoid vague or confusing UI cues (e.g., icons that change meaning depending on context).
  • Don’t force users to “figure it out”; design should allow them to proceed immediately.

• Don’t reinvent what’s already working—match mental models

  • Replacing familiar layouts/behaviors without research creates confusion.
  • Principle: recognition over recall
    • Use consistent terminology, patterns, and icons so users don’t need to relearn.
  • Keep standard components and language consistent across the product/system.

• Navigation and actions must be effortless

  • Avoid overly deep, cluttered menus or interfaces that require multiple extra clicks.
  • Provide shortcuts, searchability (when lists are long), and simple mechanisms (e.g., tabs/number patterns).

• Support recovery and accessibility

  • If an experience can fail (no internet, app crash, no battery), provide a plan B.
  • Even if the interface is “aesthetic” or app-centric, users must have a non-app path to accomplish the core task.

• Avoid deceptive design and confusing flows

  • Examples show real consequences: legal exposure, financial loss, user harm.
  • The lesson: design should not manipulate users into unintended purchases/subscriptions or trap them in unclear states.

• Respect limits, constraints, and user context

  • Good design adapts to real-world limitations (elderly users, limited connectivity, cognitive load).
  • Healthcare example: prioritize important information on the first meaningful screen, rather than dumping too much content at once.

• Reduce choice overload

  • “Choose paradox”: too many options or infinite scrolling can prevent users from deciding.
  • Provide a minimal set of top options and a sensible default (often personalized).

• Design for motivation and completion

  • Highlight “aha moments” (e.g., progress indicators).
  • Example insight: users who complete steps successfully are more likely to convert/subscribe.

• Give users ownership and control

  • Users should be able to customize or at least influence their experience.
  • If you change something, provide a way to go back—users need agency.

• Decision-makers must not design only from opinions

  • Clarissa critiques “highest-paid person” decision-making when it’s disconnected from actual user experience.
  • Use user research and experiments, not internal preferences alone.

• Survey design: mix question types, don’t over-poll

  • Yes/no questions can be biased; open questions are needed for “why.”
  • Practical approach discussed:
    • use a balanced mix of yes/no and open-ended questions
    • avoid long survey lists
    • in Uber’s practice, surveys may include incentives/bonuses and are kept short

• Culture/regions affect UX expectations

  • Different cultures scan differently (e.g., more direct vs. more context-rich).
  • Teams should adapt content/visual hierarchy accordingly, without assuming one global interaction style fits all.

• AI in design: helpful for UI/dev, limited for UX research

  • AI can assist with:
    • UI element generation/consistency
    • developer productivity
    • standardization across apps
    • generating text based on inputs/personas
    • comparing experiment results at a high level
  • But AI cannot fully replace:
    • qualitative UX research (human interviews, clustering, understanding feelings and context)
    • deeper insight into why users experience friction

• Core research principles repeated at the end

  • Don’t reinvent what’s working.
  • Match design to the real world and system expectations.
  • Ask the right questions and focus on the problem, not just the “answer.”
  • Keep users in control with visibility of system status.

---

Methodology / instruction list (detailed bullet format)

Evidence-driven UX workflow (implied methodology)

• Step 1: Identify the “expensive design mistake” via outcomes
  • Look for metrics showing users are not using features/UI as expected (e.g., not clicking a key button).
• Step 2: Use data to infer where the breakdown occurs
  • Analyze behavioral/usage data at scale (click paths, engagement, feature adoption).
• Step 3: Validate with UX research to explain “why”
  • Use interviews/surveys and qualitative insights to understand user feelings and expectations.
• Step 4: Apply fixes using proven UX patterns
  • Prefer familiar patterns that align with user mental models.
• Step 5: Test and iterate
  • Use experiments/AB testing, including by region/city when relevant.
  • Compare results and adjust based on observed behavior.

Scanning-first page/app design rules

• Make content scan-friendly by default:
  • Prioritize content with visual hierarchy (size, contrast, proximity, alignment).
  • Use bold and emphasize what matters first.
  • Break content into short paragraphs.
  • Use images/illustrations to reduce cognitive effort and support navigation.
  • Provide links or “drill-down” options for users who want more detail.

Action visibility and terminology consistency checklist

• Ensure users can immediately identify what to do:
  • Call-to-action (CTA) should be visible, large, high-contrast, and visually distinct.
  • Maintain consistent naming/terminology across screens (e.g., “Confirm pickup” should not vary arbitrarily).
• Translate the concept into implementation:
  • Keep consistent actions even if teams previously named them differently—align labels to one system.

Reduce cognitive load in information architecture

• Avoid long or unsearchable lists
  • If content is large, enable search and efficient navigation.
• Avoid unnecessary UI complexity:
  • Don’t add steps that force users to “click a bunch of times” to reach what they want.
  • Avoid unclear menu systems that open nested menus repeatedly.

Standardization and “recognition over recall”

• Reuse existing, familiar UI structures:
  • Prefer standard components and patterns rather than “new clever” layouts.
• Use icons/text that map to users’ existing expectations:
  • Avoid vague icons with ambiguous meanings.
  • Ensure icons for edit/search/zoom are contextually consistent and recognizable.

Recovery, accessibility, and fail-safe design

• Assume failures happen:
  • no internet, no battery, old users, app crashes, unsupported versions.
• Provide a self-serve fallback:
  • Include a simple alternative path (e.g., a button or non-app method to proceed).

Anti-deceptive design requirements

• Do not manipulate:
  • No hidden subscriptions/confusing buttons inside other flows.
• Ensure transparency:
  • Make the user’s intended action clear and avoid traps.

Choice overload control strategy

• Limit the initial set of options:
  • Show a minimal number of primary choices (often with a default/preference).
• Avoid infinite scrolling as a primary decision mechanism:
  • Provide structure and clarity rather than endless lists.

Survey methodology for UX feedback

• Use question types in combination:
  • Yes/no can be acceptable, but can introduce bias and may be insufficient alone.
  • Open-ended questions are needed to learn “why.”
• Keep surveys short:
  • Don’t ask very long question lists.
• Segment follow-up by context:
  • Track location, age, gender, and frequency of use to interpret results meaningfully.
• If a question is “about happiness,” also ask for reasons:
  • If users are happy, don’t over-ask; if not happy, request explanation and/or actionable hints.

AI integration boundaries (what to use AI for vs. not)

• Use AI to:
  • improve UI and developer productivity
  • standardize components
  • generate text based on structured inputs/personas
  • compare experiment outputs
• Keep humans in the loop for:
  • UX research and qualitative understanding (feelings, expectations, nuance)
  • discovering “why” behind behavioral mismatches

---

Speakers or sources featured

• Speaker: Clarissa Rodrigues (Uber; Brazil)
• Brands/organizations referenced as examples: Uber, Yamaha, New York Times, Facebook, Prime (Amazon Prime), City Bank, Apple Maps, Jeep Cherokee, Walmart, Health/healthcare systems (unnamed beyond context), Jet (juice machine example), ACP (a German company mentioned), German/Asian/American culture examples (no specific institution named)