UX design flashcards
Recall cards for the UX design track: research, synthesis, structure, visual and interaction design, accessibility, words, evaluation and design systems.
318 cards. Read the question, answer it in your head, then open the card to check.
-
Which six kinds of decision hide inside one small feature such as pausing a subscription?
Whether it is the problem worth solving now (product design), who to talk to and how (user research), the steps and states (interaction design), the words (content design), the look (UI design) and the service around it, such as the driver's list and the refund rule (service design).
From the lesson What UX design is and how teams practise it
-
How does ISO 9241-11 define usability?
As the extent to which specified users can reach specified goals with a product effectively, efficiently and with satisfaction, in a specified context of use. The users, the goal, the three measures and the context all have to be named before anyone can call a design usable.
From the lesson What UX design is and how teams practise it
-
Can the three usability measures disagree? Give the lesson's example.
Yes. In the lesson's invented test with eight people per layout, the showcase layout won on satisfaction, 4.12 against 3.50, while the plain list won on effectiveness, 100 % finished against 62.5 %, and on efficiency, 39.5 s against 73.0 s. Report the three separately: one score would hide the split.
From the lesson What UX design is and how teams practise it
-
What are the four phases of the Double Diamond, and what does each diamond do?
Discover, Define, Develop and Deliver. The first diamond widens a vague symptom into what is really happening and narrows it to one problem statement; the second widens that problem into several answers and narrows them to the one that ships.
From the lesson What UX design is and how teams practise it
-
Which four activities does ISO 9241-210 repeat until a design meets its requirements?
Understand and specify the context of use, specify the user requirements, produce design solutions, and evaluate the designs against those requirements.
From the lesson What UX design is and how teams practise it
-
What are design thinking's five modes, as Stanford's d.school lists them?
Empathize, Define, Ideate, Prototype and Test. They are modes, not a queue: the d.school's own guide plans prototyping and testing together.
From the lesson What UX design is and how teams practise it
-
Which phases does GOV.UK run a service through, and what ends discovery?
Discovery, alpha, beta and live. Discovery ends with the problem written down, and alpha is where several solutions are tried as prototypes.
From the lesson What UX design is and how teams practise it
-
When is a UX artefact, such as a clickable prototype or a handover spec, worth the time it costs?
When a decision is waiting on it. A clickable prototype earns its time when a real person needs to try the flow before it is built; a handover spec with states earns it when engineers would otherwise guess the empty, loading and error states.
From the lesson What UX design is and how teams practise it
-
Which six Gestalt principles shape what people see as one thing on a screen?
Proximity, similarity, common region, continuity and closure group items, and figure-ground decides what is in front.
From the lesson How people see: perception, attention, Gestalt
-
When can spacing alone show the groups on a screen, without borders?
When every gap between groups is at least twice every gap inside a group. Add borders or backgrounds only where the spacing cannot carry the grouping.
From the lesson How people see: perception, attention, Gestalt
-
A status pill is green for Active and amber for Paused, with no word or icon. What is wrong with it?
People who see those two colours as similar cannot tell the states apart. WCAG 2.2 criterion 1.4.1 lets colour support a meaning but never be its only visual cue, so the words Active and Paused stay on the pills.
From the lesson How people see: perception, attention, Gestalt
-
Why do people miss real content that looks like an advert, and what helps?
Banner blindness: people skip what looks like an advert or sits where adverts sit. Give the content the page's own style and put it in the main column.
From the lesson How people see: perception, attention, Gestalt
-
A change happens far from where someone is looking. What usually happens, and what helps?
They miss it: that is change blindness. Show the result where the eyes already are, and announce it as a status message.
From the lesson How people see: perception, attention, Gestalt
-
Is the F-shaped scanning pattern a rule for placing controls?
No. It describes unformatted text read quickly: two sweeps along the top, then down the left edge. Other pages and goals give other patterns, such as layer-cake, spotted, marking, bypassing and commitment reading.
From the lesson How people see: perception, attention, Gestalt
-
What do a five-second test and a blur test each tell you about a screen?
A five-second test (show a static screen for 5 s, hide it, ask what it was for) checks whether its purpose and main action come across at a glance. A blur test (blur a screenshot by 5 to 10 px, or squint) shows which elements dominate before anything is read.
From the lesson How people see: perception, attention, Gestalt
-
About how many chunks does working memory hold, and what follows for a screen?
About four, by Nelson Cowan's 2001 review. Show people what they need, such as the coupons that apply or last month's calendar with paused days marked, instead of asking them to recall it: recognition beats recall.
From the lesson Memory, decisions and movement behind UI rules
-
Should a menu be capped at seven items because of Miller's "seven plus or minus two"?
No. Miller's limits concern judging and remembering, and a menu is read, not remembered. Size it by how quickly people find and compare its items.
From the lesson Memory, decisions and movement behind UI rules
-
What does the Hick-Hyman law say about the time to choose among known options?
For n equally likely options, T = a + b log2(n + 1), so each doubling adds about the same time. When a is 201 ms and b is 149 ms per bit, the first day's fit, eight options take about 674 ms and sixteen 811 ms. A newcomer facing an unfamiliar list searches it instead.
From the lesson Memory, decisions and movement behind UI rules
-
What did the 2010 choice-overload meta-analysis find?
Across 50 experiments (63 conditions, 5,036 participants), the meta-analysis by Benjamin Scheibehenne, Rainer Greifeneder and Peter Todd found a mean effect of virtually zero, with considerable variance, and no condition that reliably produced overload. Remove what nobody needs and help people compare the rest.
From the lesson Memory, decisions and movement behind UI rules
-
How do you work out Fitts's index of difficulty for a button, and which side is W?
ID = log2(D/W + 1), with D the distance to the target's centre and W its width along the movement, in practice the rectangle's smaller side. A 48 px tall button 600 px away rates log2(13.5) = 3.75 bits; making it taller or closer lowers that, making it wider alone does not.
From the lesson Memory, decisions and movement behind UI rules
-
What does WCAG 2.2 criterion 2.5.8, Target Size (Minimum), require?
Level AA asks for pointer targets no smaller than 24 by 24 CSS pixels, with five exceptions: spacing, an equivalent control, inline targets, user agent control and essential presentation. Treat the spacing exception as a fallback, not a goal, and give primary actions far more.
From the lesson Memory, decisions and movement behind UI rules
-
What are the three response-time limits, and what feedback fits each wait?
About 0.1 s feels instantaneous, about 1 s keeps the flow of thought, and about 10 s is the limit for keeping attention on the task. Show the result itself under 0.1 s, an instant reaction up to 1 s, an indicator from 1 to 10 s, and beyond that progress with a way to stop.
From the lesson Memory, decisions and movement behind UI rules
-
Which INP counts as good, and how is it measured?
200 ms or less is good, up to 500 ms needs improvement and more is poor, read at the 75th percentile of page loads in the field, phones and desktops apart. INP reports the slowest click, tap or key press of a visit, ignoring one in every 50, from the input to the next frame painted.
From the lesson Memory, decisions and movement behind UI rules
-
When does a research question earn its place in a study?
When a decision is waiting on its answer. Write each question beside the decision it would move, and drop any question nothing rides on.
From the lesson Planning UX research: questions and methods
-
What is wrong with asking participants "Would you use feature X?"
It asks people to imagine something that does not exist and predict their own behaviour, which they do badly. Ask about something they have already done instead: the last time they wanted to skip a day of their lunch plan, what did they do?
From the lesson Planning UX research: questions and methods
-
Along which three lines do research methods differ in what they reveal?
Attitudes against behaviour (what people say against what they did), qualitative against quantitative (why against how many), and natural context against scripted tasks. Pair two methods that answer different questions, not two of the same kind.
From the lesson Planning UX research: questions and methods
-
How does discovery research differ from evaluative research?
Discovery widens a problem before any design exists; evaluative research tests a design that does. Running both in one round is how teams build the wrong thing well.
From the lesson Planning UX research: questions and methods
-
What goes into desk research before anyone is recruited?
Support logs and tickets, analytics and on-site search logs, the earlier research the organisation already holds, and a task-based competitive scan that compares the same job on two or three other services.
From the lesson Planning UX research: questions and methods
-
What can a round of five to eight research participants tell you, and what can it not?
It finds problems worth designing for, and by the fourth or fifth session the same patterns usually repeat. It cannot say how common a pattern is: five of six new subscribers missing the cut-off does not mean 83 % of all subscribers do.
From the lesson Planning UX research: questions and methods
-
Which sections does a one-page research plan hold?
Problem, research questions (3 to 5, each linked to a decision), methods, participants (screened by behaviour, access needs named), timeline, ethics (incentive, recording, retention, withdrawal) and analysis.
From the lesson Planning UX research: questions and methods
-
Why recruit research participants by past behaviour rather than demographics?
"Women aged 25 to 40 in Pune" says little about who has used a service like yours. A screener question such as "in the last four weeks, did you order lunch delivery at work?" filters for people whose behaviour matches the problem, as the GOV.UK Service Manual advises.
From the lesson Recruiting, consent and research ethics
-
What are decoy answers in a screener for?
They keep the qualifying option from being obvious. A behavioural question offers several plausible answers, only one of which qualifies, so people cannot spot which answer gets them in.
From the lesson Recruiting, consent and research ethics
-
What must a participant understand before their consent counts as informed?
Who is doing the research, its purpose, what data is collected, what will happen, how the results are used and shared, that taking part is their choice and they may stop at any point, how long the data is kept, and their rights and how to complain, plus whether the session is recorded and who is watching.
From the lesson Recruiting, consent and research ethics
-
A participant withdraws halfway through a session. What happens to the incentive and the data?
The incentive is still paid, and the recording and notes about that participant are deleted. Say "you can stop any time, without losing the incentive" during recruitment and again at the start and end of the session.
From the lesson Recruiting, consent and research ethics
-
How long should user research data be kept?
The GOV.UK Service Manual suggests two or three years as a reasonable retention period, or sooner if the data is no longer needed. Name the period on the information sheet and in the plan, and schedule the deletion.
From the lesson Recruiting, consent and research ethics
-
Is recording video "in case it proves useful later" acceptable?
No. In case is not consent, and every extra kind of data has a privacy cost the participant pays. Record only what the research questions need, such as audio when audio answers them, and pseudonymise transcripts before sharing them.
From the lesson Recruiting, consent and research ethics
-
How do you plan a research session with a participant who uses assistive technology?
Ask first about their assistive technology, support needs and preferred format for the information sheet; let them bring their own set-up, or visit them when it cannot travel; and allow extra time for consent and set-up, with breaks.
From the lesson Recruiting, consent and research ethics
-
What two habits matter more than the size of a research incentive?
The amount is printed on the information sheet before anyone agrees to take part, and payment goes out promptly, in a form the participant can use, whether or not the session produced feedback the team liked.
From the lesson Recruiting, consent and research ethics
-
Which four kinds of interview question distort the answer?
Leading questions that suggest the answer, hypothetical questions about a design that does not exist yet, double-barrelled questions that ask two things at once, and jargon questions in team words. Ask about a specific past episode instead.
From the lesson Interviews, contextual inquiry and diaries
-
What makes a probe neutral? Give two examples.
It keeps the participant's own words instead of the researcher's. "What happened next?" and "you said X, can you tell me more?" are neutral; "did you find that annoying?" colours the answer that follows.
From the lesson Interviews, contextual inquiry and diaries
-
Why should an interviewer let a silence run?
The moment a participant pauses to think is where the detail missing from the first answer arrives. The researcher's job is to try to stay quiet, and silence works as a probe of its own.
From the lesson Interviews, contextual inquiry and diaries
-
Is a discussion guide a script to read top to bottom?
No. It is a list of topics with starter questions and probes. Read line by line it turns the interview into a short survey; hold the plan lightly and bring a participant back from a side road at the next natural break.
From the lesson Interviews, contextual inquiry and diaries
-
What does contextual inquiry show that an interview room cannot, and how long is a visit?
The real place of the task: the paper on the desk, the phone that rings, the colleague the participant turns to. Visits usually run 1 to 3 hours, at most 4 one-hour or 2 three-hour visits in a day, ideally in pairs; even two or three visits pay off.
From the lesson Interviews, contextual inquiry and diaries
-
What are the three ways to observe during a contextual visit, and their trade-offs?
Silent observation keeps the task natural but leaves you guessing why; occasional questions interrupt the flow but give reasons; think-aloud gives the richest reasons and distorts the task the most. Pick one per visit and say which.
From the lesson Interviews, contextual inquiry and diaries
-
When does a diary study fit, and how is it shaped?
When the behaviour spans days, so no interview could describe it. Short, timed prompts over 7 to 14 days ask for a note, a photo or a one-line answer, through a channel the participant consented to, with the messages exported and deleted when the study ends.
From the lesson Interviews, contextual inquiry and diaries
-
What are the three columns of good session notes?
Observation (what the person did, such as scrolling the plan page for 20 seconds), quote (their exact words, with a time) and interpretation (what the team thinks it means). Interpretations stay hypotheses until analysis checks them.
From the lesson Interviews, contextual inquiry and diaries
-
Which habits carry most of the quality of a short survey question?
One idea per question, balanced wording without loaded words, an honest "don't know" option where it is a real possibility, specific and concrete periods such as "in the last 30 days", and short plain words instead of team jargon.
From the lesson Surveys and behavioural data for UX
-
Why replace an agree-disagree grid on belief statements with a forced choice?
Respondents tend to say "agree" to statements they would not have thought up themselves: acquiescence bias. Pew finds a choice between two reasonable statements, such as "I would use MealBox again" and "I would not", less prone to it.
From the lesson Surveys and behavioural data for UX
-
Which survey mode gives more honest answers on money or health, and why?
A self-administered one, such as an online survey. Phone and face-to-face modes raise social desirability bias, because people answer to the interviewer as well as the question.
From the lesson Surveys and behavioural data for UX
-
What is the margin of error at 95 % confidence for 385 and for 1,000 respondents?
About ±5 points at n = 385 and ±3 at n = 1,000, from m = 1.96 × sqrt(p(1 − p)/n) with p = 0.5. Halving a margin needs about four times the sample, because the standard error scales with 1 over the square root of n.
From the lesson Surveys and behavioural data for UX
-
Can a pop-up survey with 300 answers carry a margin of error?
Not honestly. A margin assumes a probability sample from a defined frame; a pop-up, a social post or an opt-in panel is a convenience sample that says nothing about the people who did not answer, however many did.
From the lesson Surveys and behavioural data for UX
-
How should product analytics events be named, and what stays out of them?
Small and named for what happened in the words users see, such as plan_viewed or address_entered rather than step_1_click. Account ids are usually fine; names, addresses, phone numbers and email addresses belong in a separate store with its own access controls.
From the lesson Surveys and behavioural data for UX
-
What does a log of zero-result searches tell a designer?
A short list of labels and features people expected to find and did not. It is often the cheapest source of research questions, with no recruitment at all, though like every behavioural record it says what happened, not why.
From the lesson Surveys and behavioural data for UX
-
Why did the MealBox address-step redesign need three sources?
The funnel's drop of about 21 % showed the problem was common, interviews gave three plausible reasons, and eight usability tests ranked them with the landmark field on top. Each answered a question the others could not.
From the lesson Surveys and behavioural data for UX
-
What are the six phases of thematic analysis, and how does an affinity diagram relate to them?
Familiarise, code, search for themes, review them against the notes, define them, and write up. An affinity diagram is the visual, bottom-up version of the same loop.
From the lesson Synthesis: affinity mapping and themes
-
What happens to a theme that fails to hold the notes under it?
It splits or merges. Thematic analysis is a loop, not a single pass: codes emerge from notes, themes from codes, and each theme is checked against the notes again.
From the lesson Synthesis: affinity mapping and themes
-
How do an observation, an insight and an implication differ?
An observation states what happened to the sample ("4 of 6 participants paused at the PIN code field"), an insight interprets the pattern ("people trust delivery only after they see their area is served"), and an implication says what to do ("check the PIN code before plan choice").
From the lesson Synthesis: affinity mapping and themes
-
A synthesis board has labels like "Pricing", "Delivery" and "Trust". What is wrong?
They are topics, not themes. A theme label is a sentence that says what was found; a single word is only a category.
From the lesson Synthesis: affinity mapping and themes
-
How many participants must an insight cite notes from, and why?
At least two, so a single person's note cannot carry a claim on its own, and so a sceptic can trace the insight back to the evidence. Every note keeps its participant id and, when recorded, a timestamp.
From the lesson Synthesis: affinity mapping and themes
-
Your synthesis found that 4 of 6 participants missed a field. Can you report 67 % of users?
No. Counts in qualitative research describe the sample ("4 of 6"), not the population; a rate needs a method that counts a representative sample.
From the lesson Synthesis: affinity mapping and themes
-
What two risks come with clustering research notes with an AI model?
Hallucinated themes, which sound plausible yet rest on nothing in the notes, so every theme is checked against them; and personal data exposure (OWASP LLM02, sensitive information disclosure), so names, addresses and identifying details are redacted before any text reaches a third-party model.
From the lesson Synthesis: affinity mapping and themes
-
What separates an evidence-linked persona from a stock-photo persona?
Every line cites the notes behind it and the people they come from, such as "eats at a desk at about 1 pm and buys outside if the box is late (N2, N5; P1, P3)". It describes goals, context, behaviour and concerns rather than age, hobbies and favourite brands.
From the lesson Personas, jobs to be done and scenarios
-
What is a proto-persona, and what may it be used for?
A labelled set of assumptions, used where too few people stand behind a group. It plans research; it is never the reason for a design decision.
From the lesson Personas, jobs to be done and scenarios
-
What is the form of a job story?
When a situation (what triggers the need), I want to a motivation (what the person wants to happen), so I can an outcome (why it matters to them).
From the lesson Personas, jobs to be done and scenarios
-
"When I order for my mother, I want a button to repeat last week's order, so I can finish quickly." What is wrong?
It names a mechanism, the button. Write "I want to repeat last week's order in one step": if another mechanism could replace the want and the story still holds, it sits at the right level.
From the lesson Personas, jobs to be done and scenarios
-
What does a persona help you choose, and what does a job story help you choose?
A persona helps you choose whom to design for, from patterns several participants share; a job story helps you choose what to design, from one situation. Their usual failures are a stereotype with a photo and a solution in disguise.
From the lesson Personas, jobs to be done and scenarios
-
When does a scenario become a usability-test task?
When it gives a goal and a reason with no interface words, and has a written success criterion. "You are away from Monday to Wednesday. Make sure you are not charged for those lunches" works; "Use the Pause control" gives the route away.
From the lesson Personas, jobs to be done and scenarios
-
Which domain names should a test task use for pretend addresses?
Names reserved for examples: RFC 2606 sets aside the top-level name .example, as in mealbox.example, so a made-up address in a task cannot reach a real business.
From the lesson Personas, jobs to be done and scenarios
-
In which three ways do personas mislead?
As blends or averages of different people, as lists of demographics that imply behaviour such as a device or age, and as text drafted by a language model. Anything without a source note stays an assumption to test.
From the lesson Personas, jobs to be done and scenarios
-
How does an experience map differ from a journey map?
An experience map covers the whole problem a person is trying to solve, including the parts before and after the service. A journey map is narrower: one named persona through one flow, from trigger to outcome, with actions, thoughts, feelings and pain points at each stage.
From the lesson Journey maps and service blueprints
-
What must stand behind every pain point and feeling on a journey map?
Research evidence: an interview quote, an observation note or a support ticket, with the participant id beside it. An emotion curve drawn from memory is decoration; one traced from what participants said is evidence.
From the lesson Journey maps and service blueprints
-
What should the stages along the top of a journey map name?
The part of the experience the person is in, not the screen they are on. A complaint journey starts when the resident notices the problem, not when the app opens.
From the lesson Journey maps and service blueprints
-
Which lanes does a service blueprint add below the journey?
Above the line of visibility, the frontstage the person can see; below it, the backstage actions staff take and the support systems, such as databases, SMS gateways and contractor schedules, that make those actions possible.
From the lesson Journey maps and service blueprints
-
What does the line of visibility on a service blueprint mark?
Where accountability changes, not a wall between good and bad. When an SMS is late the resident blames the service, and the blueprint shows the cause is a support system the resident has never heard of.
From the lesson Journey maps and service blueprints
-
How do current-state and future-state maps differ in what they are worth?
A current-state map comes from research. A future-state map is a hypothesis, to be checked against real behaviour before the team ships it.
From the lesson Journey maps and service blueprints
-
A confirmation reaches residents only by SMS. What should the design add?
A fallback at the channel handoff, such as an in-app confirmation straight away with SMS as the backup, because the SMS gateway can be slow or the phone out of credit. A person who moves from the app to a call should not have to explain the problem again.
From the lesson Journey maps and service blueprints
-
Which three things does a problem statement name, in one sentence?
A user specific enough for another team to picture, a need that is a goal rather than a solution, and the evidence that the need is real, citing research rather than a hunch.
From the lesson Framing problems and choosing what to solve
-
Which words most often sneak the answer into a "How might we" question?
Button, screen, popup, email, notification and modal. Rewrite the question until it could still have ten plausible answers, such as "How might we let subscribers skip a week they will be away?"
From the lesson Framing problems and choosing what to solve
-
Which beliefs does assumption mapping say to test first?
The top-left square: high importance, weak evidence, because the plan rests on them and nothing has measured them yet. Write beside each one the smallest test that could prove it wrong.
From the lesson Framing problems and choosing what to solve
-
How is a RICE score worked out, and what does its precise-looking number hide?
Reach × impact × confidence ÷ effort. Each of the four inputs is an estimate, so record who guessed each number and on what evidence; in the lesson, halving one unmeasured reach estimate dropped pause-flow from first to third.
From the lesson Framing problems and choosing what to solve
-
What three kinds of quality does the Kano model separate?
Must-be qualities, which cause dissatisfaction when missing and earn no credit when present (a served-area check); one-dimensional qualities, which raise satisfaction in step with delivery; and attractive qualities, which delight beyond what users would ask for.
From the lesson Framing problems and choosing what to solve
-
What do the letters of HEART stand for, and how is it used?
Happiness, Engagement, Adoption, Retention and Task success. Pick the categories that fit the change and walk from goal to signal to metric for each, skipping the ones the change cannot move.
From the lesson Framing problems and choosing what to solve
-
Where do accessibility and consumer-protection rules sit in a prioritisation exercise?
Above the prioritisation table, as constraints, not as features scored against the others. A design that fails WCAG 2.2 or India's dark-pattern rules is not shipped, whatever the RICE board says.
From the lesson Framing problems and choosing what to solve
-
Which four systems make up an information architecture?
The organisation scheme (how items are grouped), the labelling system (the words on groups and items), the navigation system (menus, links, breadcrumbs, sitemaps) and the search system (queries, synonyms, corrections and results), designed together.
From the lesson Information architecture: organise, label, find
-
How do exact organisation schemes differ from ambiguous ones?
Exact schemes, alphabetical, chronological or geographical, are easy to maintain but assume users know the vocabulary. Ambiguous schemes, by topic, task or audience, are harder to maintain but match how users think.
From the lesson Information architecture: organise, label, find
-
Why do labels like "Solutions" or "Resources" fail?
They make no promise about what lies behind them, so they leave no information scent. Information foraging theory (Pirolli and Card) likens choosing a link to an animal following a scent; a verb phrase in users' words, such as "Report a problem", carries more.
From the lesson Information architecture: organise, label, find
-
Which four qualities of a good GOV.UK service name work for any label?
Verbs, not nouns ("Report a noise complaint"); the words users use ("Council tax", not "Revenue and Benefits"); a task, not a technology; and no department or agency name.
From the lesson Information architecture: organise, label, find
-
How can you test candidate labels quickly, before anything is built?
With a first-click study: show a single-page wireframe with the labels and ask where people would click for a task. The clicks land in seconds, so 50 sessions fit in a day.
From the lesson Information architecture: organise, label, find
-
In what order do a content inventory, an audit and a restructure come, and what does each produce?
The inventory lists everything that exists (every URL, title, owner and date); the audit judges what survives (keep, merge, retire, rewrite); the restructure draws the sitemap last, with the survivors grouped by task. Skip the audit and the junk moves into the new structure.
From the lesson Information architecture: organise, label, find
-
Residents say "streetlight not working"; the council says "Electrical Maintenance Division". Whose words should navigation use?
The residents'. Navigation should match the users' mental model of tasks, not the organisation's chart, and first-click numbers are the evidence when a stakeholder insists on org-chart menus.
From the lesson Information architecture: organise, label, find
-
What can open, closed and hybrid card sorts each tell you?
An open sort generates groupings and vocabulary you did not have; a closed sort only reports how well your own named categories absorb the cards; a hybrid sort gives your categories and allows new ones, trading some discovery for much less coding work.
From the lesson Card sorting and tree testing
-
80 % of cards in a closed sort landed where the team put them. Does that prove the menu is good?
No. It shows people can place cards into the categories you named, not that they can find things in the structure. Propose a structure from a sort, then test findability with a tree test.
From the lesson Card sorting and tree testing
-
How many participants does a card sort need, by Tullis and Wood's evidence?
More than a usability test: the correlation with the full 168-person result was 0.75 at 5 participants, 0.90 at 15 and 0.95 at 30. Fifteen is the usual stopping point; Tullis and Wood recommended 20 to 30.
From the lesson Card sorting and tree testing
-
What does a card sort's similarity matrix hold?
One number per pair of cards: the share of participants who grouped them together. It is the only quantitative thing a sort produces.
From the lesson Card sorting and tree testing
-
Is a cut through a card-sort dendrogram a structure you can ship?
No, it is a hypothesis. Average-linkage clustering turns the matrix into a tree, and every height you cut it at gives a different structure, which must survive a test with new people and real tasks before it becomes a menu.
From the lesson Card sorting and tree testing
-
What do task success, directness and the first click measure in a tree test?
Success is the share who ended on an item that answers the task; directness is the share who never climbed back out of a branch they opened, a proxy for confidence; the first click is the branch each participant opened first. Time on task is a fourth measure.
From the lesson Card sorting and tree testing
-
Why must a tree-test task not contain the category's own label?
Participants then match words instead of finding things, and the task passes whatever the structure does. Describe the situation instead and let them choose.
From the lesson Card sorting and tree testing
-
A tree-test task has low success but high directness. What does that tell you?
A label or a placement is confidently sending people to the wrong branch. Low success with low directness means instead that no branch gave them a reason to try it.
From the lesson Card sorting and tree testing
-
What are the two halves of ideation, and why are they kept apart?
Divergent thinking makes as many options as possible with judgement held back; convergent thinking narrows them to the one or two worth building. Judged as they are spoken, half-formed ideas stop coming, and those are where good answers hide.
From the lesson Ideation, sketching and design sprints
-
Why should the first ideation round be done alone and in silence?
When making and judging happen together, the fastest and most senior voices set the direction and people anchor on early ideas. Silent individual work lets everyone arrive with options, so converging starts from a wider set.
From the lesson Ideation, sketching and design sprints
-
How does a Crazy 8s round work?
Fold a sheet into eight panels and draw one idea in each, about a minute a panel. Eight minutes is too little to polish anything, which is the point: it produces quantity.
From the lesson Ideation, sketching and design sprints
-
What question does a storyboard answer that a screen mockup does not?
Whether the idea helps in the life it lands in. Its panels show the person's experience over time, including the moments before and after any screen, such as noticing a dark street and hearing days later that it is fixed.
From the lesson Ideation, sketching and design sprints
-
How do accessibility and ethics enter an idea-scoring rubric?
As gates, not weights: an idea that excludes people or manipulates them is rejected however high it scores. In the lesson the gate removed an icon-only quick-order bar that had scored 52 of 60.
From the lesson Ideation, sketching and design sprints
-
When does a five-day design sprint fit, and what is the alternative?
A big, uncertain question worth a week, such as a new service or a decision the team keeps circling; it buys an answer from five real customer sessions. Smaller questions want a ninety-minute session with the same diverge, converge, decide shape.
From the lesson Ideation, sketching and design sprints
-
Which three rules decide whose ideas reach the shortlist in a workshop?
Individual silent work before any group talk, a visible timer on every round (to protect the divergent half, not to hurry people), and a question narrow enough to sketch against.
From the lesson Ideation, sketching and design sprints
-
What does a user flow decide, and what does a wireframe decide?
A user flow decides which states must exist and how a person moves between them, through decisions; it says nothing about where a button sits. A wireframe is the layout of one state: blocks of content and controls in place, with priority shown by size and position.
-
What is a task flow, and why is it drawn first?
A user flow narrowed to a single task with one way in, small enough to finish, so it is the first flow drawn for a feature.
-
Why draw the flow before wireframing any screen?
The flow says which screens have to exist. A screen drawn first commits you before you know the task needs it, and hides the states that appear only between screens, such as "area not served" or a declined payment.
-
Which states beside the happy path are missing from most first drafts?
Error states, which a screen shows when a step fails, and empty states, which it shows when a read returns nothing. Both say what happened and offer something to do next, and every screen needs a way out.
-
Which question does each wireframe fidelity answer?
Low fidelity, grey boxes and rough labels, answers a question about structure; mid fidelity, greyscale with the real words, a question about content; high fidelity, the finished look and behaviour, a question about how a screen is used.
-
What is a wireframe annotation for?
It tells a developer what the picture cannot, so someone who was not in the room can build the screen.
-
What should be written down before anyone builds a prototype?
The question it has to answer, such as "can a subscriber skip Thursday's delivery?". A prototype is a question in a form people can use, and that question settles what kind of prototype to make and when it has done its job.
-
Which question does each kind of prototype answer: paper, clickable, coded and Wizard of Oz?
Paper screens drawn by hand answer questions about labels and the order of steps; a clickable prototype of linked screen pictures shows whether people find a step at all; a coded HTML and CSS prototype makes keyboard, zoom and waiting behave like the real thing; Wizard of Oz puts a person behind the screen to give the answers a feature not yet built would give.
-
Is a prototype's fidelity one setting?
No. Fidelity is how closely it matches the finished thing, and it is raised one area at a time rather than with a single slider.
-
Which three traps spoil prototype sessions, whatever kind you build?
Unrealistic speed, because a prototype on the machine that made it answers instantly and hides a missing waiting state; perfect data, with short names and full lists, so no screen has to say there is nothing to show; and dead ends.
-
What is a dead end in a prototype, and when should you look for one?
A screen from which no path of links leads back to where the person started, such as a receipt with no link at all or a help link to a screen nobody built. Find them before the session by following every link, so the session is spent on the design question.
-
Which five levers decide what a screen leads with?
Size, weight, colour, contrast and space, plus position. The one primary action gets more of them than the secondary action, and many large or bold elements cancel each other out.
From the lesson Visual hierarchy, layout and grids
-
How should a secondary action look next to the primary one?
Secondary, not missing. Pale grey text at 2.54:1 on white fails the 4.5:1 of WCAG 2.2 criterion 1.4.3; the same text in the brand blue at 6.70:1 with no fill stays readable and still looks secondary.
From the lesson Visual hierarchy, layout and grids
-
What is a spacing scale, and what does it do for a layout?
A short list of allowed gaps, such as 4, 8, 12, 16, 24, 32 and 48 px, from which every margin and padding is taken. Gaps become consistent, and groups stay readable while the smallest gap between groups is at least double the largest gap within one.
From the lesson Visual hierarchy, layout and grids
-
How do you work out the column width of a column grid?
(width − 2 × margin − (columns − 1) × gutter) ÷ columns. At 1024 px with 12 columns, 32 px margins and 24 px gutters, that is (1024 − 64 − 264) ÷ 12 = 58 px, and every block starts at a column edge.
From the lesson Visual hierarchy, layout and grids
-
What do Android's window size classes describe?
The window, not the device: five classes sorted by width that set how many columns a layout uses. The class can change while the app runs.
From the lesson Visual hierarchy, layout and grids
-
CSS moves the recommended plan to the front of the screen while the code lists it fourth. Why is that a fault?
Keyboard and screen reader users meet the cards in the order of the code. WCAG 2.2 criteria 1.3.2 and 2.4.3 ask for a reading and focus order that preserves meaning, so move the card in the code instead of restyling it.
From the lesson Visual hierarchy, layout and grids
-
What does a blur test rank, and what should it be paired with?
Which block dominates a layout before any text is read. It works best beside a five-second test and a task-based test with people.
From the lesson Visual hierarchy, layout and grids
-
What is a type scale, and where should it start?
A short list of sizes in a fixed ratio, so each level of text is a clear step from the next. Start from the platform's default body size and multiply, as this track's scale does by 1.2 from 16 px; use three or four steps on one screen.
From the lesson Typography for interfaces
-
What default and minimum text sizes do browsers and iOS use?
Browsers usually default to 16 px, with no minimum set; Apple lists 17 pt as the default and 11 pt as the minimum on iOS and iPadOS, and 13 pt and 10 pt on macOS. Treat the default as a floor for body text, not a target.
From the lesson Typography for interfaces
-
Why set text sizes in rem rather than px?
A size in rem scales with the visitor's chosen default font size; a size in px does not.
From the lesson Typography for interfaces
-
How long may a line of body text be under WCAG 2.2 criterion 1.4.8?
At most 80 characters, or 40 for Chinese, Japanese and Korean, at level AAA. Cap the column in em, such as 40 em, rather than in pixels, and measure the real page when the estimate is close.
From the lesson Typography for interfaces
-
What line height suits body text, and in which unit?
At least 1.5 within paragraphs, as WCAG 1.4.8 and MDN give it, written as a unitless value so it scales with the font size.
From the lesson Typography for interfaces
-
Where do typographic settings usually break?
When a visitor enlarges text to 200 % or restyles its spacing, so text must still fit its box, and in scripts such as Devanagari, whose marks rise higher than Latin letters and need taller line boxes.
From the lesson Typography for interfaces
-
How common is low-contrast text on the web, by WebAIM's 2026 analysis?
It turned up on 83.9 % of one million home pages, the most common failure WebAIM's checker detected.
-
What is a colour role, and why do screens use roles instead of hex values?
A name for what a colour does, such as surface, on-surface, primary or error. Screens refer to the role, so one change reaches every screen; Apple defines its dynamic colours by purpose in the same way.
-
Which contrast ratios does WCAG 2.2 set for text and for interface parts?
Text needs at least 4.5:1 against its background, or 3:1 for large text (criterion 1.4.3). Component boundaries and states, and graphics that carry meaning, need 3:1 (criterion 1.4.11).
-
How is a WCAG contrast ratio calculated, and is it rounded?
(the lighter colour's relative luminance + 0.05) ÷ (the darker colour's + 0.05), with luminance from 0 at black to 1 at white, so black on white scores 21:1. It is not rounded: 4.499:1 does not meet 4.5:1.
-
A white label on #3B82F6 scores 3.68:1. How big a change does it take to pass?
Often a small one once the number is known: the slightly darker #2563EB reaches 5.17:1, above the 4.5:1 a normal-size label needs.
-
When can an icon stand without a visible word beside it?
Only when its symbol is familiar to almost everyone, such as a magnifier for search. Apple asks for a highly simplified design expressing one concept with a familiar metaphor; a stylised cart, a bin or a pair of bars means something only to people who have met it before.
-
Why does an icon-only button need an accessible name, and which criterion asks for it?
Without one, a screen reader announces it as plain "button" and a person who cannot see the symbol cannot guess what it does. WCAG success criterion 4.1.2 asks that a control's name and role can be determined by software.
-
In what order does the W3C accessible-name computation look for a control's name?
A reference to another element first, then an aria-label, then the control's content, and last of all a title attribute.
-
Is a title attribute enough to name an icon button?
No. It is a last resort that shows as a tooltip only on hover. Give the control a short action phrase, such as "Search meals", rather than a description of the picture.
-
What do signifiers, mapping, constraints and a conceptual model each do for a control?
A signifier shows what the control does (a filled button with a verb); mapping puts it next to its effect (each day's Add inside its row); a constraint stops an impossible action early (a full day refuses a third lunch); the conceptual model is the design's explanation, which should match the picture people form.
From the lesson Interaction principles, states and feedback
-
Which eight states does a control move between, and what must each have?
Default, hover, focus, pressed, loading, selected, disabled and error. Each needs a cue besides colour, and nothing a person needs may live in hover alone, because touch screens cannot hover.
From the lesson Interaction principles, states and feedback
-
A button goes grey on Friday with no explanation. What should it do instead?
Stay visible and say why, with the way out, in words beside it: "Friday is full: remove a lunch first." A grey button alone says something is wrong without saying what, so people tap it again and again.
From the lesson Interaction principles, states and feedback
-
Which WCAG 2.2 rules apply to a focus indicator?
2.4.7 (AA): a visible focus indicator; 1.4.11 (AA): an indicator outside the component contrasts 3:1 with the background it sits on; 2.4.11 (AA, new in 2.2): a focused component is never completely hidden by something the page draws on top, such as a sticky header, which scroll-padding prevents.
From the lesson Interaction principles, states and feedback
-
When is an optimistic update honest?
When it shows the change at once and takes it back with a clear message if the system refuses. The screen must never claim something the system did not do.
From the lesson Interaction principles, states and feedback
-
When does an action get an undo, and when a confirmation?
A reversible action happens at once with an undo, such as "Lunch removed. Undo"; a final one, such as cancelling a subscription, gets a confirmation that names exactly what will happen. A confirmation asked many times a day stops protecting anyone.
From the lesson Interaction principles, states and feedback
-
How long should most interface animations take, and what happens under reduced motion?
About 100 to 500 ms, with simple feedback such as a tick near 100 ms. When people set prefers-reduced-motion: reduce, motion-based animation is removed or replaced, and anything that moves by itself for more than five seconds needs a pause control in every setting (WCAG 2.2.2).
From the lesson Interaction principles, states and feedback
-
What is a question protocol, and what did it do to Ward Desk's 14-question form?
A list of each question a form asks, who uses the answer and what for, kept apart from the form. On Ward Desk it removed five questions whose answers nobody used, leaving nine.
From the lesson Forms and data entry
-
How should a form mark its mandatory and optional fields?
GOV.UK says never to mark mandatory fields with asterisks: add "(optional)" to the labels of optional ones instead. A word needs no legend, and when most fields are needed, the optional ones are the shorter list.
From the lesson Forms and data entry
-
Why is a placeholder not enough as a field's label?
It usually disappears as soon as someone types, so they lose the question while answering it. Every field needs a visible label (WCAG 3.3.2), with the format, such as "like 98765 43210", in a hint under it.
From the lesson Forms and data entry
-
Where does GOV.UK's validation pattern report errors, and when?
When the person tries to continue, not as they leave each field, and in three places: an error summary at the top linking to each field, "Error: " at the start of the page title, and a message by each field that says what happened and how to fix it.
From the lesson Forms and data entry
-
What does data entry need beyond labels and errors?
Checks that accept how real people write names, addresses, PIN codes and phone numbers; autocomplete tokens so the browser can fill fields; one-time-code fields that accept paste; password rules about length rather than character mixes; consent from the person's own tick; and upload limits stated before anyone picks a file.
From the lesson Forms and data entry
-
What are the three jobs of navigation?
Global navigation holds a service's top-level destinations on every screen; local navigation moves around inside one section; contextual navigation is the links inside the content, such as related pages.
-
How do a navigation bar and a navigation rail differ?
Both show a few top-level destinations. A bar runs along the bottom of a phone window, labelled; a rail shows the same destinations in a column at the side of a wider window. Android's default rule for adaptive navigation picks a bar if either the width or the height is compact, and a rail otherwise.
-
What do tabs and a breadcrumb trail each do?
Tabs switch between views of one thing, such as a complaint's details and its photos. A breadcrumb trail lists the pages above the current one, each a link back up.
-
Which parts make up a search pattern, one for each moment of a search?
A search field that says what can be searched, suggestions that propose whole queries while the person types, filters that narrow results and show how many remain, and a no-results page that repeats the query and offers a way on.
-
A search for "pani" finds nothing although complaints say "paani" or पानी. What is missing?
Tolerant matching: typo tolerance, which lets a small slip still match, and transliteration, which lets a word written in another script or spelling find the same thing.
-
In which six places can an interface message appear?
An inline message beside its field or list, a notification banner across the top of the page, a toast at the screen's edge that leaves by itself, a popover beside the control that asked for it, a bottom sheet rising from the bottom of a phone screen, and a modal dialog that blocks the screen until it is answered.
-
How does a modal dialog differ from a non-modal one?
A modal dialog blocks the page behind it until its question is answered, and an alert dialog is its urgent kind; a non-modal dialog leaves the page usable behind it.
-
A list is empty. What should its empty state say?
Which of the two reasons applies: nothing yet, or nothing that matches. An empty state is a message, not a blank area that looks broken.
-
Besides empty, which states does a screen need to say something in?
Loading (what is coming), partial (what arrived and what did not), offline (the copy kept on the device, with its age), success (what was done) and error (what went wrong, what is still safe and how to try again).
-
In which three ways can a message fail to reach someone, and which WCAG criteria cover them?
It leaves before it is read or used while nothing else offers the same thing (2.2.1, Timing Adjustable); it is silent to a screen reader (4.1.3, Status Messages); or it moves by itself with no way to pause it (2.2.2, Pause, Stop, Hide).
-
Why did the three-second Undo toast in the lesson fail twice?
It was gone in three seconds while nothing else offered the undo, and a screen reader was never told it had appeared. The inline error and the dialog reached everyone because they stayed until dealt with.
-
What changes about a target on a touch screen?
A fingertip covers many pixels and nothing can hover, so the size of a target and the room around it decide what gets hit.
-
Lunches can be reordered only by dragging a small handle. What must be added?
A way to do the same with single taps or clicks, such as Move up and Move down buttons. WCAG 2.2 asks this of anything done by dragging; gestures that follow a path or need two fingers also need a single-tap alternative.
-
What must be true of a single-key keyboard shortcut?
One of three things holds: people can switch it off, people can remap it, or it acts only while its control has focus.
-
What is the label-in-name rule, and who needs it?
A control's accessible name contains its visible label, so people who use voice control can press it by saying what they see. A button that says Save week but is named Submit plan fails it.
-
Which inputs does a design most often forget, and why do they fail together with the keyboard?
Voice control, which acts on the words a person says, and switch access, which highlights the screen one item at a time. Both lean on the keyboard's rules: every action reachable, and a focus order that keeps the page's meaning.
-
What is the difference between a responsive layout and an adaptive layout?
A responsive layout keeps one arrangement and stretches or shrinks it continuously within a range of widths. An adaptive layout switches to a different arrangement when the width crosses a breakpoint. Most products do both: they flow within each range and switch between ranges.
-
Where should a breakpoint go?
Where the content breaks, such as a line that grows uncomfortably long or a row that no longer fits. Popular device widths change with every release cycle, so they make poor breakpoints.
-
Which window size classes does Android define by width?
Compact below 600 dp, medium from 600 dp, expanded from 840 dp, large from 1200 dp and extra-large from 1600 dp. The class decides the frame of the screen, such as a navigation bar or a rail.
-
How do Apple's size classes describe the space an app has?
Each dimension is compact or regular, set from the device, the window and multitasking. Apple asks apps to lay out by size class rather than by device type or orientation: an iPhone Duo's outer display is compact width and its inner display regular width.
-
What are Android's three canonical layouts?
List-detail shows a list and the chosen item side by side at expanded widths and one at a time below that. Supporting pane gives the main content about two thirds of the window and the rest to a second pane whose content means nothing alone, such as filters. Feed puts items of one kind in a grid that grows from one column to many.
-
Across iOS, Android and the web, what stays the same and what follows each platform?
The content, the words and the brand stay the same. Navigation, sheets and Back behave as each platform's own apps do: iPhone apps use a tab bar, Android apps a navigation bar, and Back is left to the system or the browser.
-
What does WCAG 2.2's Reflow criterion ask, and why 320 CSS px?
Success criterion 1.4.10, level AA, asks that content stay usable in a viewport 320 CSS px wide without scrolling in two directions, the width a 1280 px window leaves at 400 % zoom. Maps, diagrams, video, games and data tables may scroll inside their own box.
-
What do WCAG's Resize Text and Orientation criteria ask?
Resize Text, 1.4.4, asks that text grow to 200 % without loss of content or function. Orientation, 1.3.4, forbids locking a page to one orientation, portrait or landscape, unless that one is essential, as for a bank cheque or a piano app.
-
How does Indian digit grouping differ from Western grouping?
Western grouping separates every three digits: 1,234,567. Indian grouping puts the first separator after three digits and the rest after every two: 12,34,567, that is 12 lakh, 34 thousand and 567. CLDR writes the pattern as #,##,##0.###.
-
How should an interface produce a rupee amount for an Indian reader?
Format it with Intl.NumberFormat in the reader's chosen locale and the currency style with INR, which reads the grouping, digits and currency symbol from the locale's data. Building it from strings, or with a comma every three digits, gets the grouping wrong; check the output of every locale you ship.
-
Where does compact notation such as 1.3Cr belong, and where does it not?
On a chart, not on a bill. Compact notation turns 1,25,00,000 into 1.3Cr, rounded to two significant digits, so the exact amount is gone.
-
Which digits do hi-IN and mr-IN use, and how do Indian and US short dates differ?
hi-IN keeps Latin digits unless asked for Devanagari ones with -u-nu-deva, while mr-IN uses Devanagari digits from the start. The short date puts the day first in en-IN and the month first in en-US.
-
How should a language switcher name its languages?
Each in its own name and script, as an endonym: हिन्दी, मराठी, English, which Intl.DisplayNames returns. A list written only in English makes a reader of Marathi alone look for a word in a script they do not read.
-
What do WCAG 3.1.1 and 3.1.2 ask of a page in several languages?
3.1.1 Language of Page, level A, asks for a lang attribute on the html element, which every element inherits, so screen readers load the right pronunciation rules. 3.1.2 Language of Parts, level AA, gives a phrase in another language its own attribute. Keep tags short: Hindi is hi.
-
What does a 75th-percentile connection in India mean for page weight?
About 6.2 Mbit/s down, 2.1 Mbit/s up and an 85 ms round trip, and a quarter of people have slower ones. At 6.2 Mbit/s every megabyte needs at least 1.29 seconds just to arrive.
-
Which phone and settings should a team test on for Indian customers?
A low-cost Android, since one estimate puts devices at the 75th percentile close to three years old, with Data Saver on, which blocks background data on metered networks, and with TalkBack, Android's screen reader.
-
What does it mean to see disability as a mismatch?
The problem sits where a person meets a design, not in the person. A barrier appears where a design demands something a person cannot give, so the design is where the barrier can be removed, and every design decision adds or removes a mismatch.
From the lesson Inclusive design and disability
-
How many people live with a significant disability, by the World Health Organization's estimate?
About 1.3 billion, roughly one person in six or 16 % of the world's population. The figure counts only significant disability; temporary injuries and situations that take an ability away for a while reach many more people.
From the lesson Inclusive design and disability
-
What is the persona spectrum?
A way of widening one demand into permanent, temporary and situational cases. A device designed for a person with one arm also serves someone with an injured wrist and a new parent holding a baby, and one change removes the barrier for all three.
From the lesson Inclusive design and disability
-
Why is a flow only as open as its least open step?
A person blocked at one step cannot finish, however open the other steps are. So every step people must pass is checked against every demand it makes.
From the lesson Inclusive design and disability
-
How is WCAG 2.2 organised, and what does level AA mean?
Each requirement is a testable success criterion at level A, AA or AAA. WCAG 2.2 has 86 criteria: 31 at A, 24 at AA and 31 at AAA. Level AA means meeting every A and every AA criterion.
From the lesson Inclusive design and disability
-
Why design to WCAG 2.2 AA when the Indian government-site guidelines and the US rule name 2.1 AA?
The laws lag the standard, and a page that meets WCAG 2.2 meets 2.1 and 2.0 as well. A product designed to 2.2 AA therefore meets every one of those references at once.
From the lesson Inclusive design and disability
-
Which Indian law and standards set the accessibility target for digital services?
The Rights of Persons with Disabilities Act makes the Central Government responsible for accessibility rules. Rule 15 of its rules points to GIGW 3.0, the guidelines for government websites, and to IS 17802 for ICT products; the Supreme Court has since directed the Union to set mandatory rules, so check what has been notified.
From the lesson Inclusive design and disability
-
How is inclusion planned into a project instead of audited at the end?
Disabled people take part in every research round, the WCAG criteria are checked on sketches, manual and automated checks run with each new feature, and the audit is the last check rather than the only one. A barrier costs least where it is decided: a drag-only picker found on paper costs a new sketch.
From the lesson Inclusive design and disability
-
Where are most WCAG 2.2 AA criteria met or missed?
In the design file, by a size, a colour, a place, a word or a state. Quantity buttons 20 px wide, a savings line marked only by green, a pale field border and a missing focus indicator each fail a criterion before any code exists.
-
What do WCAG's colour criteria 1.4.1, 1.4.3 and 1.4.11 ask?
1.4.1 Use of Color, level A: meaning carried by more than colour. 1.4.3 Contrast (Minimum), AA: 4.5:1 for text. 1.4.11 Non-text Contrast, AA: 3:1 for graphics.
-
Which six A and AA criteria are new in WCAG 2.2?
Focus Not Obscured (Minimum) at 2.4.11, Dragging Movements at 2.5.7, Target Size (Minimum) at 2.5.8, Consistent Help at 3.2.6, Redundant Entry at 3.3.7 and Accessible Authentication (Minimum) at 3.3.8. All six describe decisions made in a design file.
-
What do 2.4.7 Focus Visible and 2.4.11 Focus Not Obscured (Minimum) ask of a design?
A focus state that is drawn for every control, and that nothing covers, such as a sticky bar over the focused element.
-
What must stay the same from page to page under 3.2.3, 3.2.4 and 3.2.6?
Navigation keeps one order, each function keeps one name, and help sits in one place. So a design check covers whole flows as well as single screens.
-
What do WCAG's error and sign-in criteria ask of a form?
Errors are identified and come with their fix (3.3.1, 3.3.3), fields have labels or instructions (3.3.2), people do not retype what they already gave (3.3.7 Redundant Entry), and signing in sets no memory test (3.3.8 Accessible Authentication).
-
How many of WCAG 2.2's 55 A and AA criteria can axe-core's rules look at?
axe-core 4.13.0's default rules are tagged with 19. Its WCAG 2.2 rule, off by default, adds 2.5.8 Target Size and its experimental rules two more, for 22; the other 33 have no rule at all, and a tagged criterion is only partly covered.
-
Can a clean automated scan show that a screen is accessible?
No. It shows only that the failures its rules look for were not found. Focus hidden by a sticky bar, help that moves and 20 px buttons pass a default scan, so run the scan on every build and check the rest in the design, with a keyboard and with a screen reader.
-
Why does a heading drawn only as large bold text fail screen reader users?
Assistive tech reports what the code says, not what the picture shows. In WebAIM's survey of screen reader users, 67.8 % look for information on a long page by moving through its headings, so text that only looks like a heading gives them nothing to move by.
-
What is an accessibility annotation?
A note on the design that a developer turns into code and a tester checks. It writes down what each part means, so the meaning no longer depends on anyone reading the picture the same way.
-
Which layers does an accessibility annotation cover on every screen?
Landmark regions such as banner, navigation, main and contentinfo; heading levels; the reading order and the order Tab moves in; an accessible name, role and state for every control; alt text for each image; and the status messages a screen reader must announce.
-
Where should the names in an accessibility annotation come from?
From visible text and native HTML first, as W3C's ARIA Authoring Practices advise. A good annotation names the element to use, such as a button or a heading, before any attribute.
-
What is plain language?
Writing that the intended reader can find, understand and act on the first time. It is written for what the reader knows and needs to do, which may be technical, so it is not dumbing text down.
From the lesson UX writing: plain language, voice, structure
-
Which four moves do most of the work of plain language?
Use the reader's words, say who does what in the active voice, put the action first, and cut what the reader skips. Then check that every fact survives the cuts: a shorter pause notice that drops the 48-hour deadline gets people charged.
From the lesson UX writing: plain language, voice, structure
-
What is the difference between a product's voice and its tone?
The voice is the product's character in words and stays the same everywhere. The tone is how it speaks right now, and the reader's moment decides it: never playful when someone may be worried or money is at stake.
From the lesson UX writing: plain language, voice, structure
-
Along which four dimensions does Nielsen Norman Group describe tone?
Humour, formality, respectfulness and enthusiasm, each a scale rather than a switch. GOV.UK adds that services avoid humorous error messages, because people often use a service under stress.
From the lesson UX writing: plain language, voice, structure
-
What is front-loading, and why does it matter?
Putting the word the reader is scanning for at the start of a heading, link or button. People scan rather than read, and a scanning reader usually takes in only about the first 11 characters of a link, about two words.
From the lesson UX writing: plain language, voice, structure
-
What does a readability formula such as Flesch reading ease measure?
Only how long the sentences are and how long the words are; higher is easier, and 10-word sentences of 1.5 syllables a word score 69.8. It cannot see order, meaning or accuracy: the same words in reverse order score exactly the same.
From the lesson UX writing: plain language, voice, structure
-
Where are readability scores unreliable?
On one-line strings, whose scores run off both ends of the scale, and on other languages, because the weights were fitted to English text. A score means nothing for a Hindi string.
From the lesson UX writing: plain language, voice, structure
-
What measures whether interface text is clear?
Readers. Paraphrase testing asks a person to say in their own words what a short text means, and usability testing covers longer content and forms; test early, revise and test again, at least twice. A flag from a checker is a question for a person, not a verdict.
From the lesson UX writing: plain language, voice, structure
-
What should a button label say?
What will happen when it is pressed: a verb first and its object after, in the words of the question on the page. When a dialog asks Pause your plan?, the buttons are Pause plan and Keep plan.
-
Why avoid button labels such as Submit, OK, Yes and Cancel?
They fit every button, so people must read the rest of the page to know what they agreed to, and Cancel can even hide two opposite outcomes, such as cancelling the pause or cancelling the plan. In a dialog both buttons say what they do.
-
What should link text say?
Where the link goes, so it makes sense on its own. W3C's writing tips rule out click here and read more; Read how pausing works says where it leads.
-
What makes a good error message, by GOV.UK's guidance?
It uses the words of the field's label, says what went wrong and gives the fix: an instruction such as Enter your mobile number, or a rule such as Name must be 60 characters or fewer. It has no blame or filler words such as invalid, please, sorry and oops, no technical codes, and nothing so general it fits any field.
-
Where does an error message appear on a form that follows GOV.UK's pattern?
Beside its field, and repeated exactly in an error summary that heads the page and links to the field. A hidden Error: before the message tells screen reader users what it is.
-
Which problem gets a page instead of an error message?
One the person cannot fix in their answer, such as an area MealBox does not deliver to. The page explains the problem and what to do next.
-
What should a confirmation, an empty screen or first-use text say?
What happened and what to do next: the same rule an error message follows, applied after an action.
-
What makes a usability test task worth running?
A short scenario the participant can picture, with a reason to want the thing done and a success criterion written next to it. It names the person, the goal and the situation but never a button or menu: Use Schedule to set Friday only tests only whether a label is readable.
-
What is a think-aloud session, and which two kinds of report does it use?
The participant says what they are thinking while they do the task. Ericsson and Simon separated concurrent reports, said during the work, from retrospective reports, given after it.
-
A participant falls silent for twenty seconds. What does the moderator do?
Let the silence work, since it is often reading, deciding or recovering from a surprise. If they are still stuck, ask a neutral prompt such as What are you thinking about? or What would you try next on your own?, never one that points at a control.
-
Which settings can a usability test use, and what decides between them?
In-person moderated, remote moderated or remote unmoderated. The choice follows what the round must produce: live probes, scale, help when things break, and recruiting reach.
-
What does the problem-discovery formula say about how many participants are enough?
A round of n people finds a problem that affects a share p of users with probability 1 minus (1 minus p) to the power n. Five catch 84.4 % of problems at p = 0.31 but only 41.0 % of a one-in-ten problem, which needs 19 participants to reach 85 %.
-
What did Laura Faulkner's study show about rounds of five users?
From 60 tested users, some random sets of five found 99 % of the problems while others found only 55 %, and the weakest set of ten still found 80 %. Any one small round can vary a lot.
-
What do the points of Nielsen Norman Group's 0 to 4 severity scale mean?
0 is not a usability problem at all, 1 is cosmetic, 2 is minor, 3 is major and 4 is a usability catastrophe to fix before release.
-
Which three dimensions set a usability problem's severity?
Frequency, impact and persistence: how often users meet it, how badly it hurts and whether it comes back on every attempt. Writing each down lets raters disagree about one dimension, and severity follows what the round saw, not what the fix will cost.
-
What is a usability inspection?
A way to find usability problems without participants, through people who know what to look for. Heuristic evaluations and cognitive walkthroughs are both quick and cheap, and neither replaces watching real users.
-
How does a heuristic evaluation work?
A few evaluators each check every screen alone against a short list of rules of thumb, such as Nielsen's ten usability heuristics, and then pool what they found.
-
Why does Nielsen Norman Group ask for three to five evaluators working independently?
Each person misses some problems, and different people miss different ones, so the pooled list holds what each evaluator alone would miss.
-
What turns a broken heuristic into a finding someone can act on?
Writing what a person meets before the rule it breaks: Slot: nothing changes for 8 seconds after Pay is pressed, so people may pay twice (visibility of system status). A heuristic number alone names a rule, not what goes wrong.
-
Which lists besides Nielsen's ten heuristics can an evaluation use?
Shneiderman's eight golden rules and the interaction principles of ISO 9241-110, plus extra rules for the product's own category.
-
What is a cognitive walkthrough, and when does it suit?
A group agrees on the user, often a persona, and the correct steps of one task, then asks at each step whether someone using the product for the first time would get through it. It suits learnability questions: a flow new to its users, or a task done too rarely to remember.
-
Why do inspections come before a usability test, never instead of it?
Neither watches a user, so both judge only what the team can imagine: they cannot show which words customers expect or how a busy kitchen or a weak network changes the task, and some of what they flag no customer meets. Inspecting first keeps the test's few participants for what only they can show.
-
Four of five testers finished a task. What success rate does that show?
80 % as a single number, but the true rate it stands for could sit anywhere from about 36 % to 98 %. Every number from a small usability test is an estimate and needs a range before anyone acts on it.
-
Why is time on task summarised with a geometric mean rather than a plain average?
Time on task is skewed by a few slow people, so a plain average describes the typical person badly.
-
What are SUS, SEQ and UMUX-Lite?
The System Usability Scale turns ten questionnaire answers into one score from 0 to 100, which is not a percentage. The Single Ease Question is one seven-point rating of how hard a task felt, asked as each task ends. UMUX-Lite is a two-item questionnaire that tracks the SUS and fits inside a longer survey.
-
How is the Net Promoter Score computed, and what does it measure?
The percentage of promoters, who rate their likelihood to recommend 9 or 10, minus the percentage of detractors, who rate it 0 to 6. It measures loyalty, not usability.
-
Why does a small test's success rate need an adjusted-Wald interval?
It stays honest at small samples, where the textbook formula can run past 100 %. Like every confidence interval, it says which true values could still have produced what you saw.
-
A design's true success rate is 80 %. How often do four or more of five testers finish?
About 74 % of the time, so the same design can show 100 %, 80 % or 60 % depending only on who turned up that day.
-
Which design question does each of an A/B test, a usability test and product analytics answer?
A usability test answers why people struggle. Product analytics answers where and how often people leave, never why. An A/B test answers whether a change worked: which of two versions does better on a chosen number, and by how much.
-
Why does an A/B test send each visitor at random to one version?
So that chance alone decides who sees what, and a difference larger than chance can be put down to the change. It still cannot say why the number moved, and it needs a built version and enough visitors, often thousands for a small change.
-
What does a funnel from product analytics show, and what can it not show?
Where people leave, from events the live product logs: of 1,000 visitors to a plan page, 380 choose a plan and 100 pay. It never shows why, and a change seen after a launch can have another cause, such as a festival week.
-
Why is no single method enough for a design decision?
Quantitative methods such as A/B tests and analytics answer what happened, and qualitative methods answer why. The authors of the HEART framework advise triangulating large-scale metrics with usability studies and field studies.
-
What are an experiment's primary metric and guardrail metrics?
The primary metric is the one number the decision rests on. Guardrail metrics are numbers that must not get worse while it moves, such as how many people cancel in their first week.
-
What do peeking, a novelty effect and a sample-ratio mismatch do to an A/B test?
Peeking, reading the result daily and stopping on a good day, invents winners. A novelty effect or change aversion fades as people get used to a change. A sample-ratio mismatch means the split itself failed, so no result can be read.
-
What are the five parts of a design system?
Foundations (colour, type, spacing and layout decisions every screen shares), components (reusable parts, each specified, built and tested), patterns (proven ways to do one task), guidance (when to use each part, how to word it and how it must behave) and governance (who decides what goes in and how it changes).
From the lesson Design systems: components, docs, governance
-
What are the three ways to get a design system, and which should a small team choose?
Adopting an existing one, adapting one or creating your own; the more custom the system, the more time and money it costs. A small team adopts a public system, themes it with its own colours, type and spacing, and builds only the components no public system offers.
From the lesson Design systems: components, docs, governance
-
Why not fork a public design system to change its colours?
Every fix the original team ships then has to be merged by hand into a copy that drifts, and the original's testing no longer covers the parts you changed. Change colours, type and spacing through the system's theming and leave its components as they are.
From the lesson Design systems: components, docs, governance
-
What does a component spec cover?
Its anatomy (the named parts), variants, states with visual and accessibility notes, content limits as numbers anyone can test, keyboard behaviour, and what a screen reader hears: the role, the accessible name and any state, such as aria-pressed on a toggle.
From the lesson Design systems: components, docs, governance
-
What makes usage guidance for a component useful?
It says when to use the component and when not to, with do and don't pairs that each give a reason. For a chip: do use it to filter the menu by diet, because it turns one filter on or off; don't use it to open another page, because a link says where it goes.
From the lesson Design systems: components, docs, governance
-
How does GOV.UK's contribution model decide what enters its design system?
Anyone may propose a component or pattern. The proposal must be useful, with evidence that many teams need it, and unique. Before publication it must also be usable, shown in research that includes disabled people, consistent with the system and versatile enough for other services.
From the lesson Design systems: components, docs, governance
-
What do the three numbers of a Semantic Versioning release promise?
MAJOR.MINOR.PATCH: the major number rises for incompatible changes, the minor for additions that keep compatibility and the patch for compatible fixes. For a component, a removed variant, a renamed property or a new default height is incompatible, because teams might have to change their screens.
From the lesson Design systems: components, docs, governance
-
How does a design system remove a part without surprising anyone?
It deprecates the part first, in a minor release, so at least one version carries the warning, and removes it in the next major release, listed as a breaking change. A changelog that names each replacement keeps the upgrade small.
From the lesson Design systems: components, docs, governance
-
What problem do design tokens solve?
They name each design decision once, in a file that design tools and code both read, so changing a value means changing one place. A brand green written 14 different ways in 230 places is missed somewhere when it changes.
-
What does a token look like in the Design Tokens Community Group format?
A JSON object with a $value, placed in groups. A group may set $type for every token inside it, and the format forbids guessing a type from the value; a token may also carry $description, $deprecated and $extensions. Files end in .tokens or .tokens.json.
-
Which token types does the format define?
Basic types such as color, dimension (a number with px or rem), duration (a number with ms or s), fontFamily, fontWeight, cubicBezier and number. Shadows, borders, transitions, gradients and typography are composite types built from them.
-
What is a token alias, and when is it an error?
A value such as {color.green.600} that refers to another token; chains of aliases are followed to the end. Aliases that point at each other form a loop, and every token in a loop is an error, as is an alias to a name that does not exist.
-
What do primitive, semantic and component tokens each name?
A primitive token, such as color.green.600, holds an explicit value and says which colours exist. A semantic token, such as color.action.primary, names what a value is for. A component token, such as button.primary.background, names where it is used.
-
Why do component tokens point at semantic tokens, never at primitives?
A theme changes the semantic tier, and every component that points at it follows. A component token that skips to a primitive keeps its light value in every theme.
-
What does theming mean with design tokens?
Giving a few semantic names different values for each theme, such as light, dark or high contrast, while every component keeps the names it already uses.
-
What do a resolver document, a transform and a contrast matrix do?
A resolver document combines token sets with modifiers, such as a theme with light, dark and high-contrast contexts, into one set of values per choice. A transform turns each resolved token into code, such as a CSS custom property. A contrast matrix measures every text and surface pair in every theme.
-
How does a design critique differ from a design review?
A critique improves the work against its stated goals and ends with ranked problems and questions that have owners. A design review decides whether the work goes ahead, with any conditions. Each needs its own meeting.
From the lesson Critiques, design reviews and research readouts
-
What are the three roles in a design critique?
The presenter, whose work it is; the critics; and a facilitator, who sets the scope, keeps time and keeps the talk on the goals. Rotating the facilitator stops one person always steering the outcome.
From the lesson Critiques, design reviews and research readouts
-
Which four parts make a critique comment something a designer can act on?
A tag (a question, an observation or a suggestion), the goal it is about, evidence that says why, and a severity from 0 to 4. Make the header pop more has none of them.
From the lesson Critiques, design reviews and research readouts
-
Which openings keep the tags of feedback honest?
I notice… for an observation, I wonder… for a question and What if… for a suggestion. They also stop comments from turning into commands.
From the lesson Critiques, design reviews and research readouts
-
In what order should a designer act on the comments from a critique?
By severity against the goals: the 3s and 4s first, then an answer to the rest, including the ones you decide not to act on. Working through comments in the order they arrived gives a cosmetic suggestion the same day's work as a catastrophe.
From the lesson Critiques, design reviews and research readouts
-
How does a remote critique work?
A written round of about a day, in which everyone comments on the frames, tagged and tied to a goal; the facilitator then groups the comments by goal and severity and merges duplicates; a call of about 20 minutes takes only disagreements and comments rated 3 or 4.
From the lesson Critiques, design reviews and research readouts
-
How is a research readout put together?
It opens with the decision the research informs, shows each finding with its evidence at its true size, such as 6 of 8 rather than 75 %, and closes with recommendations that each have an owner and a sprint, because a recommendation nobody owns is a wish.
From the lesson Critiques, design reviews and research readouts
-
What does a decision log record, and why does it keep replaced decisions?
One entry per decision: what was decided, why, the alternatives considered, the evidence, who decided and its status. Replaced entries stay, because the reason a future team needs most is why something was once done differently.
From the lesson Critiques, design reviews and research readouts
-
What are handoff and design QA?
Handoff is the work of getting a design built as intended: the file, plus the specs that say how it behaves, the tokens and assets it uses, and the conversations around them. Design QA is the check afterwards, comparing the build with the design at agreed widths, states and themes.
-
When does a handoff that works start?
Over the first sketches, with the engineers, when an answer still costs a sentence: where the data comes from, how long a name can be, which states the server can return. It continues through the build instead of being a file sent at the end.
-
What does a handoff package hold?
The specs, the design tokens every value comes from, the assets (icons exported as SVG, photos in modern formats with a fallback) and the accessibility annotations. A screen is ready when a checklist covering its states, tokens, assets and annotations is complete.
-
Why are icons handed off as SVG?
SVG scales to any size, which MDN calls ideal for icons. An icon exported as PNG is a gap a handoff checklist should catch.
-
What can an inspection panel such as Figma's Dev Mode show, and what can it not?
It shows one frame's layout, spacing and colours, the variables it uses and generated code, and lets people download assets. It cannot show what a frame does not draw: how the layout changes between widths, which state uses which colour, long or translated text, focus order and what a screen reader hears.
-
How should engineers treat code generated by an inspection tool?
As a hint: it describes one frame at one size. A component from the design system already carries its states and its keyboard behaviour.
-
Besides its states, what behaviour must a screen's spec settle?
Content limits and truncation rules, which say how long each text may grow and what happens beyond that, responsive rules and localisation, which say how the screen changes across widths and languages, and its analytics events.
-
What is dual-track working?
A discovery track, which turns ideas into backlog items that have been checked, runs beside a delivery track, which turns those items into software people can use. Research and design run a sprint or so ahead of the build.
-
Which habit marks designers who thrive in agile teams, by Nielsen Norman Group's interviews?
They work at least one step ahead of the sprint. The GOV.UK Service Manual keeps research going after launch as well, to assess how people experience the live service and to test changes.
-
What does a design hypothesis name?
The change, the people, the outcome expected and the signal that will show it, so the change becomes a bet that can be checked.
-
What does an outcome roadmap order, and why does it matter?
The changes in behaviour a team aims for, rather than a list of features. A squad that ships 11 features with no outcome written down cannot say which feature was meant to move a number, and a skip rate going from 18.4 % to 18.1 % stays inside the month-to-month noise.
-
How should a team report the impact of UX work?
As a measured change with its uncertainty, such as a rate with its 95 % interval, never as a borrowed return-on-investment multiplier.
-
What are ResearchOps and DesignOps?
The shared routines and tools that keep research and design repeatable as a team grows.
-
What separates persuasion from a dark pattern?
A dark pattern works only while the customer cannot see the facts behind it; persuasion still works when they see everything. A stock line that matches the kitchen's real count is persuasion, and a countdown that restarts on every reload is a dark pattern.
-
What do India's guidelines on dark patterns define, and how many patterns do they name?
The Central Consumer Protection Authority defines a dark pattern as a deceptive design practice meant to mislead or trick people into doing what they did not intend. Its guidelines name 13 patterns, from false urgency to rogue malware, and cover platforms offering goods or services in India, advertisers and sellers.
-
What are basket sneaking and drip pricing?
Basket sneaking adds items or payments without consent, such as a dessert ticked before checkout. Drip pricing shows part of the price only later, such as a delivery fee added on the payment screen.
-
What does rule 4(9) of India's E-Commerce Rules ask of a checkout?
A purchase counts only when the customer agrees by an explicit, affirmative action of their own, so every paid add-on starts with its box empty.
-
What does the EU's Digital Services Act ask of an online platform's design?
That platforms design their services so that they do not deceive or manipulate users, or impair their ability to make informed decisions of their own. One honest design meets both the EU's rule and India's.
-
What does each finding of a dark-pattern self-audit record?
The screen, the element, the pattern, the evidence that proves it and a fair fix. The audit also clears what only looks suspicious, and India's amended rules make the self-audit yearly.
-
What makes a subscription fair rather than a trap?
It is as easy to leave as to join, with no phone call and no required reason, and a reminder before each renewal gives the amount and a link to cancel. A hidden, long or confusing cancellation is a subscription trap, and a trial that turns paid without notice is SaaS billing.
-
What guardrail should an engagement goal carry?
A measure that rises when people are being tricked, such as cancellations started against cancellations completed, complaints about unexpected charges, refunds and disputed payments, and opt-outs after each new prompt.
-
When does AI fit a product problem?
When a wrong answer is easy to notice and cheap to fix, such as a suggested lunch the customer can see and swap. Where the result must be exact or predictable, such as sorting a menu by price, a plain rule is as good or better and simpler to build, explain and debug.
-
What is a hallucination in an AI feature?
Content that seems plausible but is made up, which a model unsure of an answer may produce.
-
What should an AI feature tell people before they first rely on it?
What it can do and how well it can do it, in words they can act on, such as a note that suggestions can miss foods you avoid, so each dish should be checked before ordering. People then rely on it as much as it deserves, and a promise of a perfect week every week sets an expectation it cannot keep.
-
A feature gets 3 of 20 suggested weeks wrong. How likely is a customer who takes four suggested weeks to meet a wrong one?
About 48 %, which is 1 minus 0.85 to the power 4. A modest error rate per answer adds up quickly over repeated use.
-
How do Microsoft's guidelines for human-AI interaction arrange their 18 rules?
In four phases: when people first meet the feature, while they use it, when it is wrong, and over time. Each phase is a state the screen must show.
-
Which states of an AI feature need design as carefully as the finished result?
The waiting, the partial, the wrong and the refused states, because people meet all of them.
-
How should an AI feature handle a hard constraint such as an allergy?
It is checked by a rule before any dish is shown.
-
What is the rule for using AI tools in UX work?
The tool drafts, a person checks, and each kind of draft has a check of its own. Fluent text is not evidence: a model unsure of an answer may produce content that seems plausible but is made up, and present it convincingly.
-
How is an AI tool's summary of interview transcripts checked?
Each claim gets traced back to a quote, to the person who said it and to a count of how many others said the same.
-
Which four wording slips does a check of drafted interview questions look for?
A leading question, a double-barrelled question that asks two things at once, a hypothetical question about an imagined future, and a question that invites a plain yes. Questions about something that happened pass.
-
What does a clean run of an interview-question check show?
Only that its word patterns found none of those slips, not that the guide is good. A researcher still reads it for order, coverage and tone.
-
How should a team treat the answers of synthetic users?
As questions to ask real people, never as findings.
-
How is participants' personal data protected when AI tools are used?
Use only tools the participants consented to, and mask personal details before any text is sent out.
-
What do prompt-to-prototype output and the final report each need?
Prototype output is for exploring directions: its markup is checked and it is tested with people before any of it ships. The report names the steps where an AI tool helped and how each output was checked.
-
What should the first screen of a case study carry?
The problem, your role and the result, with where the number comes from. Readers decide from the first screen, so a list of process stages or a persona before the problem tells them the method but not why the work mattered; the process belongs in the middle.
From the lesson UX portfolio case studies
-
How many projects belong in a UX portfolio, and how are they chosen?
Two to four that together show process, judgement and outcomes and differ from each other, such as one weighted towards research, one towards interaction or visual design and one where a constraint forced a hard trade-off. At least one should have had real users.
From the lesson UX portfolio case studies
-
What are the seven parts of a case study?
Problem, role, constraints, evidence, decisions and trade-offs, outcome and reflection, each answering a question the reader has. A short part is better than a missing one.
From the lesson UX portfolio case studies
-
When is an outcome in a case study evidence rather than a claim?
When it says what was measured, over what period and from where, such as from the subscription records over two months after launch. A number without its source is a claim.
From the lesson UX portfolio case studies
-
What must be taken out of research evidence before a case study is published?
Names, faces, voices and telling details from participants' quotes, figures nobody agreed to share, real customer data in screens and notes with personal data. Quotes are used only with consent.
From the lesson UX portfolio case studies
-
How does a case study stay honest about team work, AI help and confidential work?
Name each person's part and use I for yours; say what an AI tool did and what you checked or changed; for work under a non-disclosure agreement, ask permission first and, if it is refused, describe it in general terms with figures as ranges and no screens.
From the lesson UX portfolio case studies
-
What must a self-initiated redesign of a real product say?
That it was self-initiated and whether you tested it with anyone, that you are not affiliated with the company and used public information only, and it uses no logo in a way that suggests the company endorsed it.
From the lesson UX portfolio case studies
-
Which checks does a case-study page itself need to pass?
WCAG 2.2 AA: text at 4.5:1, alt text that follows each image's purpose, empty alt for decoration, and headings that never skip a level. For speed, at the 75th percentile: LCP within 2.5 s, INP of 200 ms or less and CLS of 0.1 or less, so the first-screen image is not lazy-loaded.
From the lesson UX portfolio case studies
-
What does every format of design interview test?
Whether people who were not in the room can follow your reasoning and judge it.
-
What does each common design interview format let a panel see?
A portfolio presentation shows how you tell one project; an app critique, how you judge a product you did not design; a whiteboard challenge, how you frame and explore a new problem aloud; a take-home, how you work alone to a limit; a research or content exercise, a specialist skill; behavioural questions, stories about past work.
-
What should you ask the recruiter before preparing for design interviews?
Which formats your interviews will include and how long each one lasts. Employers combine the formats in their own ways, and preparing for the wrong one wastes the days before.
-
How do you budget the words of a 15-minute portfolio talk?
Time yourself reading a 300-word paragraph aloud at your presenting pace. At 2 minutes 30 seconds you speak 120 words a minute, so 14 minutes of talk holds about 1,680 words, and one minute stays spare. Use your own pace, not an average.
-
Which part of a portfolio talk deserves the longest slot?
The decisions and their trade-offs, because a panel can read about research methods anywhere but can hear your reasons only from you.
-
What does a whiteboard design challenge reward?
A visible method more than a polished screen: the questions you ask before drawing, what you assume aloud, a direction chosen with a reason, and how you answer pushback. Going back to the flow when a direction breaks a constraint is part of the method.
-
How are the seven moves of this lesson's whiteboard method arranged?
Three moves frame the problem before any drawing, the fourth sketches the user flow, and the last three choose a direction, weigh its trade-offs and close with a short summary.