Zeeshan Khalid
Book a 30-min callBook a call

Design practice

Design methods guide

63 methods for every phase of a product design project, from understanding users to validating what you shipped. Each one says when to use it, how to run it, what you get and the mistake to avoid, how AI can help and a note from practice. Not sure where to start? Let's plan your project first.

Let's plan your project

Pick where you are, the time you have and who can take part. You get a short list of methods to run, in order.

Time for each method
Who can take part

Your starting plan: 6 methods, in order

The methods most worth running first for a whole project with this time and team. It's a starting point, not the whole toolkit: swap in others from the 63 methods below when your project needs them.

  1. Understand

    Analytics review

    At the start of a project, to learn what is already known and to aim research well.

    2–5 daysOn your own

  2. Understand

    Journey mapping

    You need to see the whole experience, not one screen, and where expectation and reality part.

    2–5 daysWith a few teammates

  3. Define

    Problem statement

    Before ideation, so everyone solves the same problem.

    1–2 hoursWith a few teammates

  4. Ideate

    Impact and effort matrix

    Choosing what to do first from a list of ideas.

    30–60 minutesWith a few teammates

  5. Execute

    User flows and edge cases

    Before detailed design, and before handoff.

    1–3 days per flowWith a few teammates

  6. Validate

    Heuristic evaluation

    Quickly finding obvious problems before testing with users, or when users are hard to reach.

    1–3 daysWith a few teammates

Browse all 63 methods

Choose a phase, search or filter by time and who takes part. Open a method for the steps, what you get and how AI can help.

63 methods in five phases

Understand

Learn who the users are, what they are trying to do and what gets in their way, and how the business makes money and measures success. Mix numbers (what happens) with conversations (why it happens).

Quantitative research

  • A fixed set of questions sent to many people, to measure how common an attitude, behaviour or problem is.

    1–3 weeksOn your own

    Understand · Quantitative research

    Survey

    A fixed set of questions sent to many people, to measure how common an attitude, behaviour or problem is.

    • 1–3 weeks
    • On your own
    • Starter

    When to use it

    You need numbers (how many, how often, how satisfied), or to check whether something heard in interviews applies widely.

    How to run it

    1. Write the decision the survey should inform.
    2. Keep it short (5–10 minutes) with closed questions and one or two open ones.
    3. Pilot it with 3–5 people.
    4. Send it to the right sample.
    5. Analyse by segment, not just totals.

    People: 1 designer or researcher

    What you get

    Percentages and segments, with quotes from the open questions.

    Watch out

    Leading or double questions ("How easy and fast was…?"), and treating what people say they do as what they do.

    With AI

    Ask AI to review a draft for leading, double or vague questions and to suggest answer scales; let it group open answers into themes. Check the themes against the raw answers before you report them.

    From practice
    At PaysWell, research across 200+ user responses fed the roadmap of a survey platform for 100K+ users, and lifted satisfaction from 3.2 to 4.5 stars.
  • Measuring how many people complete each step of a key flow (sign-up, checkout, booking) and where they drop off.

    1–3 daysOn your own

    Understand · Quantitative research

    Funnel analysis

    Measuring how many people complete each step of a key flow (sign-up, checkout, booking) and where they drop off.

    • 1–3 days
    • On your own
    • Starter

    When to use it

    You know the goal of a flow and need to find the step that loses the most people.

    How to run it

    1. Define the steps and the event for each.
    2. Pick a time range and segment (device, channel, new or returning).
    3. Find the biggest drop.
    4. Watch sessions or interview users at that step to learn why.

    People: 1 designer with analytics access

    What you get

    A step-by-step conversion chart and the one or two steps worth fixing first.

    Watch out

    A funnel shows where, not why; pair it with qualitative research.

    With AI

    Analytics tools with AI assistants can find the biggest drop and the segments behind it in seconds. Use it to point you to the step; the reason still comes from watching sessions or talking to users.

    From practice
    At Educative, a metrics culture on Mixpanel and Amplitude made funnels part of every design decision; hypothesis-driven tests improved conversion funnels by 35%.
  • Reading existing product, web and marketing data (traffic, behaviour, search terms, support tickets) before new research.

    2–5 daysOn your own

    Understand · Quantitative research

    Analytics review

    Reading existing product, web and marketing data (traffic, behaviour, search terms, support tickets) before new research.

    • 2–5 days
    • On your own
    • Starter

    When to use it

    At the start of a project, to learn what is already known and to aim research well.

    How to run it

    1. List the questions.
    2. Gather dashboards, reports and support data.
    3. Note patterns and gaps.
    4. Turn gaps into research questions.

    People: 1 designer, help from analytics or marketing

    What you get

    A one-page summary of what the data shows and what it doesn't.

    Watch out

    Vanity numbers (page views) instead of behaviour tied to goals.

    With AI

    Paste exported reports (never personal data) and ask AI for the patterns, the gaps and the questions worth researching. Verify every number it quotes against the source dashboard.

    From practice
    At GreatPeople, analytics alongside research and usability testing shaped decisions for performance tools used by 50K+ people a day.

Qualitative research

  • One-to-one conversations about people's goals, behaviour and problems in their own words.

    1–3 weeksOn your own

    Understand · Qualitative research

    User interviews

    One-to-one conversations about people's goals, behaviour and problems in their own words.

    • 1–3 weeks
    • On your own
    • Starter

    When to use it

    You need to understand why: motivations, context and workarounds.

    How to run it

    1. Write 5–8 open questions about past behaviour ("Tell me about the last time…").
    2. Recruit 5–8 people per user group.
    3. Listen more than talk, ask "why" and "show me".
    4. Note quotes and observations straight after.
    5. Synthesise with affinity mapping.

    People: 1–2 people, 5–8 participants per group

    What you get

    Themes, quotes and needs per user group.

    Watch out

    Asking people to predict the future or design the solution ("Would you use…?").

    With AI

    Use AI to draft a neutral discussion guide, transcribe sessions and pull first-pass themes and quotes. Do the interviewing yourself, get consent for recordings, and read the transcripts: AI summaries smooth over the surprises that matter.

    From practice
    At GreatPeople, 30+ interviews, usability studies and stakeholder workshops shaped the roadmap for performance tools used by 50K+ people a day.
  • Watching and asking people while they do their real work in their real setting.

    1–3 weeksOn your own

    Understand · Qualitative research

    Contextual inquiry

    Watching and asking people while they do their real work in their real setting.

    • 1–3 weeks
    • On your own
    • Intermediate

    When to use it

    The work is complex, physical or full of workarounds people don't think to mention.

    How to run it

    1. Agree a session at the person's place of work (or a screen share).
    2. Observe, then ask about what you see.
    3. Note tools, interruptions and workarounds.
    4. Confirm your reading with them at the end.

    People: 1–2 people, 4–8 participants

    What you get

    Work models, workarounds and needs people would not report in an interview.

    Watch out

    Turning it into an interview; the value is in what you see.

    With AI

    AI helps afterwards: turn field notes and photos into a tidy work model, or a first list of workarounds. It can't observe for you, so keep the visit itself human.

    From practice
    For the Pakistan Ministry of Defence ERP and MES platform, research with 30+ operational staff across procurement, inventory and production planning shaped the information architecture.

    Origin: Hugh Beyer and Karen Holtzblatt, Contextual Design (1998).

  • A moderated discussion with 6–8 people about a topic, product or idea.

    1–2 weeksWith a few teammates

    Understand · Qualitative research

    Focus group

    A moderated discussion with 6–8 people about a topic, product or idea.

    • 1–2 weeks
    • With a few teammates
    • Intermediate

    When to use it

    Early, to hear language, opinions and reactions to a concept; not to test usability.

    How to run it

    1. Recruit people with something in common.
    2. Prepare a discussion guide.
    3. Moderate so every voice is heard.
    4. Analyse themes, not counts.

    People: 1 moderator, 1 note-taker, 6–8 participants per group

    What you get

    Attitudes, language and reactions, with quotes.

    Watch out

    One strong voice sways the group; what people say in a group is not what they do alone. Never use it to judge whether something is easy to use.

    With AI

    Transcribe and summarise the discussion with AI, then check who said what: AI summaries often merge one strong voice into "the group".

    From practice
    At J&R Marketing, working with clients and marketing teams across 20+ products in healthcare, sports and security put customers' language at the start of each design.
  • Participants record their experiences over days or weeks, as they happen.

    2–6 weeksOn your own

    Understand · Qualitative research

    Diary study

    Participants record their experiences over days or weeks, as they happen.

    • 2–6 weeks
    • On your own
    • Intermediate

    When to use it

    The behaviour unfolds over time (habits, multi-step journeys, occasional tasks).

    How to run it

    1. Define what and when to record.
    2. Give a simple tool (app, form, messages).
    3. Prompt and check in regularly.
    4. Run a short interview at the end.

    People: 1 researcher, 8–15 participants

    What you get

    Real behaviour over time, with moments of friction and their triggers.

    Watch out

    Drop-off; keep entries short and thank people often.

    With AI

    AI can send reminders, tag entries by theme and flag entries that need a follow-up question. Keep the reading human; diary entries are personal and full of context.

    From practice
    On Nitro League, a racing game with 85% weekly retention, how players came back week to week mattered as much as any single session.
  • A one-page summary of a user group in four quadrants: says, thinks, does, feels.

    1–2 hoursGroup workshop

    Understand · Qualitative research

    Empathy mapping

    A one-page summary of a user group in four quadrants: says, thinks, does, feels.

    • 1–2 hours
    • Group workshop
    • Starter

    When to use it

    After research, to align the team on what you learned; as a starting point before personas.

    How to run it

    1. Draw the four quadrants around the user.
    2. Fill them with research evidence (sticky notes).
    3. Mark contradictions (says vs does).
    4. List needs underneath.

    People: The team

    What you get

    A shared picture of a user group and its needs.

    Watch out

    Filling it with assumptions instead of research; label anything unverified.

    With AI

    Give AI your research notes and ask it to sort them into says, thinks, does and feels as a starting draft. Never let it fill the quadrants from imagination; every note should trace back to research.

    From practice
    At Educative, user research across developer personas informed feature priorities for a learning platform used by 100K+ developers.

    Origin: Dave Gray, XPLANE.

  • Participants group topics (on cards) in a way that makes sense to them, and name the groups.

    1–2 weeksOn your own

    Understand · Qualitative research

    Card sorting

    Participants group topics (on cards) in a way that makes sense to them, and name the groups.

    • 1–2 weeks
    • On your own
    • Intermediate

    When to use it

    Designing or fixing navigation and information architecture.

    How to run it

    1. Write 30–60 cards of real content.
    2. Choose open (they name groups), closed (you give groups) or hybrid.
    3. Run it online or in person.
    4. Analyse which cards are grouped together.

    People: 1 designer, about 15–30 participants for an online open sort

    What you get

    Groupings and labels that match people's mental models.

    Watch out

    Cards that share a word get grouped by the word, not the meaning.

    With AI

    Use AI to draft the card list from your content inventory and to cluster results. Don't use AI "participants": the point is how real people group things.

    From practice
    For the Pakistan Ministry of Defence ERP and MES platform, information architecture spanned procurement, inventory, production planning and quality control, the kind of scope card sorting is made for.
  • A short comic of a user's situation, need and how a product helps, frame by frame.

    1–3 hoursOn your own

    Understand · Qualitative research

    Storyboarding

    A short comic of a user's situation, need and how a product helps, frame by frame.

    • 1–3 hours
    • On your own
    • Starter

    When to use it

    Explaining a scenario to stakeholders, or exploring a concept before designing screens.

    How to run it

    1. Pick one scenario and person.
    2. Write 4–8 steps.
    3. Sketch each (stick figures are fine).
    4. Add the feeling at each step.

    People: 1 designer or the team

    What you get

    A scenario everyone can follow.

    Watch out

    Showing the interface too early; focus on context and emotion.

    With AI

    Image generators make quick storyboard frames from short prompts. Keep them rough and consistent, and label them as concepts, not designs.

    From practice
    On PaidAnyday, a wallet for same-day payouts, the experience spanned employers and employees, so scenarios had to show both sides of the payment.
  • A collage of images, type, colour and references that sets a visual direction.

    1–2 daysOn your own

    Understand · Qualitative research

    Moodboard

    A collage of images, type, colour and references that sets a visual direction.

    • 1–2 days
    • On your own
    • Starter

    When to use it

    Before visual design, to agree the tone with stakeholders or clients.

    How to run it

    1. Collect references for each brand quality.
    2. Group them into 2–3 directions.
    3. Review and choose one.
    4. Note what to keep from the others.

    People: 1 designer

    What you get

    An agreed visual direction.

    Watch out

    Approving a look that doesn't fit the brand or accessibility (check contrast early).

    With AI

    AI image and search tools speed up collecting references and generating variations of a direction. Check licences before reusing any image, and check colour contrast early.

    From practice
    For flydubai's move from low-cost to premium, the visual direction had to change across five websites at once.

Holistic view

  • A timeline of a user's steps, thoughts, feelings and pain points while reaching a goal, across every channel.

    2–5 daysWith a few teammates

    Understand · Holistic view

    Journey mapping

    A timeline of a user's steps, thoughts, feelings and pain points while reaching a goal, across every channel.

    • 2–5 days
    • With a few teammates
    • Intermediate

    When to use it

    You need to see the whole experience, not one screen, and where expectation and reality part.

    How to run it

    1. Choose one persona and goal.
    2. List stages and steps from research.
    3. Add actions, thoughts, feelings and touchpoints.
    4. Mark pain points and opportunities.

    People: Designer with the team

    What you get

    A map of the current experience with the moments worth fixing.

    Watch out

    A map drawn from assumptions; base each step on evidence.

    With AI

    Ask AI to turn interview notes into a first draft of stages, actions and pain points, then correct it with the team. Mark each step with its evidence so the map stays honest.

    From practice
    At J&R Marketing, journey mapping was part of the full design lifecycle across 20+ products, which lifted average website conversion by 35%.
  • A diagram of all the people, organisations, systems and channels around a product, and how value and information flow between them.

    1–2 daysWith a few teammates

    Understand · Holistic view

    Ecosystem map

    A diagram of all the people, organisations, systems and channels around a product, and how value and information flow between them.

    • 1–2 days
    • With a few teammates
    • Intermediate

    When to use it

    Products with many parties (marketplaces, government services, enterprise platforms).

    How to run it

    1. Put the user in the centre.
    2. Add actors and systems around them.
    3. Draw the flows (money, data, services).
    4. Mark gaps and dependencies.

    People: Designer with product and business

    What you get

    One picture of who is involved and where things break between them.

    Watch out

    Making it so detailed nobody can read it; keep one level of detail.

    With AI

    AI can draft a first list of actors and flows from documents and interviews, and generate a diagram. Review it with people who run the service; AI misses informal links.

    From practice
    Saudi Energy's Vision 2030 platform served 35M+ users across B2C, B2B and B2E, with many parties and systems behind each journey.

Business

  • Conversations with the people who fund, run, sell or support the product.

    1–2 weeksOn your own

    Understand · Business

    Stakeholder interviews

    Conversations with the people who fund, run, sell or support the product.

    • 1–2 weeks
    • On your own
    • Starter

    When to use it

    At the start, to learn goals, constraints, success measures and politics.

    How to run it

    1. List stakeholders by influence.
    2. Ask about goals, risks, metrics and history.
    3. Note where answers conflict.
    4. Play the summary back to them.

    People: 1–2 people, 5–10 stakeholders

    What you get

    Business goals, constraints and the conflicts to resolve.

    Watch out

    Treating the highest-paid opinion as the requirement.

    With AI

    Use AI to prepare question sets per role and to compare answers across stakeholders, highlighting where goals conflict. Share the summary with them to correct it.

    From practice
    On Saudi Energy's Vision 2030 platform, a design enablement framework for 300+ product teams depended on early alignment with those teams' leads.
  • A map of a service from the customer's view (front stage) and everything behind it (back stage, systems, policies) that makes it work.

    3–5 daysWith a few teammates

    Understand · Business

    Service blueprint

    A map of a service from the customer's view (front stage) and everything behind it (back stage, systems, policies) that makes it work.

    • 3–5 days
    • With a few teammates
    • Advanced

    When to use it

    The experience depends on people and processes behind the screen.

    How to run it

    1. Start with the customer's steps.
    2. Add front-stage actions, back-stage actions and supporting systems in rows.
    3. Draw the line of visibility.
    4. Mark failure points and waiting times.

    People: Designer with operations

    What you get

    Where the service breaks behind the interface, and who owns the fix.

    Watch out

    Leaving out the back stage; that is where most problems are.

    With AI

    AI can draft the back-stage rows from process documents and spot missing handovers. Validate it with operations staff, who know where the real waiting happens.

    From practice
    For AbbVie, a workflow platform for 500+ scientists and production staff covered task automation, approval routing and compliance tracking, much of it behind the screen; approvals became 50% faster.

    Origin: G. Lynn Shostack, "Designing Services That Deliver", Harvard Business Review (1984).

  • Describing the progress a person is trying to make in a situation ("When…, I want to…, so I can…"), independent of any product.

    1–2 weeksOn your own

    Understand · Business

    Jobs to be done

    Describing the progress a person is trying to make in a situation ("When…, I want to…, so I can…"), independent of any product.

    • 1–2 weeks
    • On your own
    • Intermediate

    When to use it

    You need to understand why people choose or switch products, beyond demographics.

    How to run it

    1. Interview recent switchers about the moment they decided.
    2. Write job statements.
    3. Group them.
    4. Use them to judge ideas.

    People: 1–2 people, 8–12 participants

    What you get

    Job statements that guide features and messaging.

    Watch out

    Writing jobs that are really features ("I want a dashboard").

    With AI

    Ask AI to rewrite interview quotes as job statements in the "When…, I want to…, so I can…" form, and to group similar jobs. Discard any job it can't link to a quote.

    From practice
    On PaidAnyday, the job was clear: get paid the same day; the wallet reached 85% adoption among participating employers.

    Origin: Popularised by Clayton Christensen (Competing Against Luck, 2016) and Tony Ulwick.

Define

Turn what you learned into a clear problem, agreed with the team: who has it, why it matters to them and to the business, and how you will know it's solved.

Vision and strategy

  • A canvas matching a customer profile (jobs, pains, gains) with what the product offers (products, pain relievers, gain creators).

    2–4 hoursGroup workshop

    Define · Vision and strategy

    Value proposition canvas

    A canvas matching a customer profile (jobs, pains, gains) with what the product offers (products, pain relievers, gain creators).

    • 2–4 hours
    • Group workshop
    • Starter

    When to use it

    Shaping a new product or feature, or checking that one still fits its users.

    How to run it

    1. Fill the customer side from research.
    2. Fill the product side.
    3. Draw the matches.
    4. Note what is unmatched.

    People: Product, design, business

    What you get

    A clear statement of why the product matters to a group of users.

    Watch out

    Filling the customer side without research.

    With AI

    AI is useful for challenging the canvas: ask it which pains are unaddressed or which claims competitors already make. The customer side must still come from research.

    From practice
    For flydubai's move to premium, the design system had to carry a new value proposition across five brands' websites.

    Origin: Alexander Osterwalder and others, Strategyzer (Value Proposition Design, 2014).

  • A 2×2 of internal strengths and weaknesses and external opportunities and threats.

    1–2 hoursGroup workshop

    Define · Vision and strategy

    SWOT analysis

    A 2×2 of internal strengths and weaknesses and external opportunities and threats.

    • 1–2 hours
    • Group workshop
    • Starter

    When to use it

    Agreeing where a product stands before choosing a direction.

    How to run it

    1. Gather evidence (research, competitors, data).
    2. Fill the four boxes.
    3. Pick the few items that matter.
    4. Turn them into actions.

    People: Team and stakeholders

    What you get

    A shared view of the situation and 3–5 actions.

    Watch out

    Long lists with no decisions; end with actions.

    With AI

    Ask AI for a first SWOT from your research and market notes, then argue with it. Its threats and opportunities are generic unless you give it real evidence.

    From practice
    At Tkxel, establishing design standards across 30+ concurrent client projects started from where each product stood.
  • A structured session at the start of a project to agree goals, scope, roles and risks.

    2–4 hoursGroup workshop

    Define · Vision and strategy

    Kick-off workshop

    A structured session at the start of a project to agree goals, scope, roles and risks.

    • 2–4 hours
    • Group workshop
    • Starter

    When to use it

    Day one of any project with more than a few people.

    How to run it

    1. Share goals and what's known.
    2. Run hopes and fears (sticky notes, then group).
    3. Agree roles (a RACI helps) and ways of working.
    4. List open questions and owners.

    People: The whole project team

    What you get

    Agreed goals, roles, risks and next steps.

    Watch out

    A presentation instead of a workshop; everyone should write and speak.

    With AI

    AI helps prepare: an agenda, hopes-and-fears prompts and a first RACI draft from the project brief. After the session, it can turn the board into actions with owners.

    From practice
    At flydubai, RACI-based governance for the design system made ownership clear across five websites and a team of four UX and UI designers.

User needs

  • Grouping research notes by theme to find patterns.

    2–4 hours per round of researchGroup workshop

    Define · User needs

    Affinity mapping

    Grouping research notes by theme to find patterns.

    • 2–4 hours per round of research
    • Group workshop
    • Starter

    When to use it

    After interviews, tests or workshops, before deciding anything.

    How to run it

    1. One observation per sticky note.
    2. Group silently by similarity.
    3. Name each group with a statement.
    4. Rank the groups.

    People: The team

    What you get

    Themes and insights backed by evidence.

    Watch out

    Naming groups with one word ("Navigation") instead of an insight ("People can't find past orders").

    With AI

    Whiteboard tools can cluster sticky notes with AI in seconds. Treat the clusters as a first pass: rename them as insights and move the notes that don't fit.

    From practice
    At SecurEnvoy, findings from usability testing with 50+ enterprise IT administrators were grouped into the login problems worth fixing first; login-related tickets fell by 30%.

    Origin: The KJ method, Jiro Kawakita (1960s).

  • Short, research-based profiles of the main user groups: goals, behaviour, context and frustrations.

    2–5 days after researchOn your own

    Define · User needs

    Personas

    Short, research-based profiles of the main user groups: goals, behaviour, context and frustrations.

    • 2–5 days after research
    • On your own
    • Intermediate

    When to use it

    The team needs a shared, specific picture of who it designs for.

    How to run it

    1. Find behaviour patterns in research.
    2. Write one persona per pattern (3–5 total).
    3. Lead with goals and behaviour, not demographics.
    4. Use them in reviews and decisions.

    People: Designer or researcher

    What you get

    3–5 personas the team refers to.

    Watch out

    Invented personas, or ones full of hobbies that change no decisions.

    With AI

    AI can draft persona write-ups from your research summaries. Don't use invented "synthetic users" in place of research; they describe the average and miss your actual users.

    From practice
    At Educative, research across developer personas informed feature prioritisation and content strategy for 100K+ developers.

    Origin: Alan Cooper (The Inmates Are Running the Asylum, 1999).

  • Breaking a user's task into its steps, decisions and dependencies.

    1–3 daysOn your own

    Define · User needs

    Task analysis

    Breaking a user's task into its steps, decisions and dependencies.

    • 1–3 days
    • On your own
    • Intermediate

    When to use it

    Designing or improving a complex task (forms, configuration, operations work).

    How to run it

    1. Observe or ask how the task is done today.
    2. List steps and sub-steps.
    3. Note decisions, information needed and errors.
    4. Find steps to remove or support.

    People: 1 designer

    What you get

    A task model and the steps to simplify.

    Watch out

    Modelling how the task should be done instead of how it is done.

    With AI

    Give AI an observation transcript and ask it to break the task into steps and decisions. Check it with the people who do the work; AI tidies away the messy steps you need to see.

    From practice
    For AbbVie, reducing cognitive load in high-stakes research and manufacturing work cut manual task processing time by 40%.
  • Describing how users think something works, and comparing it with how the product works.

    1–2 weeksOn your own

    Define · User needs

    Mental models

    Describing how users think something works, and comparing it with how the product works.

    • 1–2 weeks
    • On your own
    • Advanced

    When to use it

    People keep making the same mistake, or a new concept is hard to explain.

    How to run it

    1. Ask users to explain how they think it works (talk aloud, drawings).
    2. Note their terms and steps.
    3. Compare with the product.
    4. Change the design or the words to match.

    People: 1 researcher, 5–8 participants

    What you get

    The gaps between user thinking and the product, with fixes.

    Watch out

    Expecting users to adopt the system's model; usually the design should move.

    With AI

    AI can compare users' explanations with how the system works and list the mismatches. The explanations must come from real users, in their own words.

    From practice
    On Virtua, an NFT marketplace, buying across several blockchain networks had to make sense to newcomers; 92% of first-time buyers completed their task.

    Origin: Indi Young, Mental Models (2008), for the mapping technique.

Business opportunity

  • A structured review of competitors and alternatives: what they offer, how they work, where they are weak.

    3–5 daysOn your own

    Define · Business opportunity

    Competitive analysis

    A structured review of competitors and alternatives: what they offer, how they work, where they are weak.

    • 3–5 days
    • On your own
    • Starter

    When to use it

    Entering a market, setting priorities, or before a redesign.

    How to run it

    1. List direct and indirect competitors (including spreadsheets and doing nothing).
    2. Choose the criteria.
    3. Review each the same way (tasks, screenshots, scores).
    4. Find gaps and conventions to keep.

    People: 1 designer

    What you get

    A comparison table with the opportunities and the conventions users expect.

    Watch out

    Copying features without knowing why they exist.

    With AI

    Use AI to collect public features, pricing and reviews of competitors quickly, and to fill a comparison table. Check claims on the competitors' own sites; AI summaries go out of date.

    From practice
    At PaysWell, competitive analysis and usability studies shaped the gamified reward mechanics of a survey platform for 100K+ users.
  • Choosing measurable goals for the user experience in five areas: Happiness, Engagement, Adoption, Retention and Task success, each broken into goals, signals and metrics.

    2–4 hoursWith a few teammates

    Define · Business opportunity

    Goals and metrics (HEART)

    Choosing measurable goals for the user experience in five areas: Happiness, Engagement, Adoption, Retention and Task success, each broken into goals, signals and metrics.

    • 2–4 hours
    • With a few teammates
    • Intermediate

    When to use it

    Before designing, so the team agrees what success looks like.

    How to run it

    1. Pick the 2–3 areas that matter for this project.
    2. Write a goal for each.
    3. Find a signal (behaviour or attitude) that shows progress.
    4. Turn it into a metric with a baseline.

    People: Product, design, data

    What you get

    A short table of goals, signals and metrics.

    Watch out

    Tracking everything; pick a few metrics you will act on.

    With AI

    Ask AI to suggest signals and metrics for each goal, and which your analytics already track. Keep the final list short and owned by the team.

    From practice
    At Horizontal Digital, agreeing measurable design success metrics with product, engineering and leadership reduced cross-functional friction by 60%.

    Origin: Kerry Rodden, Hilary Hutchinson and Xin Fu, Google (CHI 2010).

  • Objectives (a qualitative goal) with 2–4 Key results (measurable outcomes) for a period, usually a quarter.

    1–2 hours a quarterWith a few teammates

    Define · Business opportunity

    OKRs

    Objectives (a qualitative goal) with 2–4 Key results (measurable outcomes) for a period, usually a quarter.

    • 1–2 hours a quarter
    • With a few teammates
    • Starter

    When to use it

    Linking design work to company or team goals.

    How to run it

    1. Write an objective people care about.
    2. Add key results that measure outcomes, not tasks.
    3. Review progress weekly.
    4. Score at the end and learn.

    People: Team leads

    What you get

    Shared goals the design work is judged against.

    Watch out

    Key results that are tasks ("Launch the redesign") instead of outcomes.

    With AI

    AI is good at spotting key results that are really tasks, and at rewording them as outcomes. The objectives themselves need the team's judgement.

    From practice
    At Horizontal Digital, design work was tied to business objectives with product, engineering and executive leadership.

    Origin: Andy Grove at Intel; popularised by John Doerr (Measure What Matters, 2018).

Problem analysis

  • Asking "why" repeatedly (about five times) to get from a symptom to a root cause.

    30–60 minutesGroup workshop

    Define · Problem analysis

    Five whys

    Asking "why" repeatedly (about five times) to get from a symptom to a root cause.

    • 30–60 minutes
    • Group workshop
    • Starter

    When to use it

    A problem keeps coming back, or the obvious fix seems too shallow.

    How to run it

    1. State the problem.
    2. Ask why it happens.
    3. Ask why of each answer.
    4. Stop at a cause you can act on.
    5. Check it with data.

    People: The team

    What you get

    A root cause and a fix that targets it.

    Watch out

    One chain of whys for a problem with several causes; branch where needed.

    With AI

    Use AI as a sparring partner: ask it for other possible causes at each step. Confirm the chain with data, not with the AI's confidence.

    From practice
    The Pakistan Ministry of Defence ERP cut data-entry errors by 40%, the kind of result that comes from fixing causes, not symptoms.

    Origin: Sakichi Toyoda; part of the Toyota Production System.

  • Framing a problem with seven questions: what, why, where, when, who, how and how much.

    30–60 minutesWith a few teammates

    Define · Problem analysis

    5W2H

    Framing a problem with seven questions: what, why, where, when, who, how and how much.

    • 30–60 minutes
    • With a few teammates
    • Starter

    When to use it

    A problem or request is vague and needs to be made specific.

    How to run it

    1. Write the problem.
    2. Answer each question with evidence.
    3. Mark unknowns.
    4. Rewrite the problem in one sentence.

    People: Designer with product

    What you get

    A complete, specific problem description.

    Watch out

    Guessing answers; unknowns become research questions.

    With AI

    Paste a vague request and ask AI to fill the seven questions and list what's unknown. The unknowns are the useful part: they become questions for the requester.

    From practice
    At Tkxel, structured design governance and stakeholder alignment kept delivery consistent across 30+ concurrent projects.
  • Listing what must be true for an idea to work, then ranking assumptions by importance and evidence.

    1–2 hoursGroup workshop

    Define · Problem analysis

    Assumption mapping

    Listing what must be true for an idea to work, then ranking assumptions by importance and evidence.

    • 1–2 hours
    • Group workshop
    • Starter

    When to use it

    Before investing in a solution, to choose what to test first.

    How to run it

    1. List assumptions (users, value, feasibility, business).
    2. Place them on a 2×2: important / unimportant, known / unknown.
    3. Test the important unknowns first.

    People: The team

    What you get

    The riskiest assumptions and how to test each.

    Watch out

    Listing facts as assumptions to feel safe.

    With AI

    Ask AI to list hidden assumptions in your idea, by type (user, value, feasibility, business). Then rank them yourselves; only the team knows what evidence it has.

    From practice
    On Nitro League, wallet connection and asset management were the riskiest parts of onboarding, so they were simplified first; the game raised $5M in 20 days.

    Origin: David J. Bland (Testing Business Ideas, 2019).

Problem statement

  • One or two sentences that say who has the problem, what it is, why it matters (to them and the business) and how you'll know it's solved. Includes the current and ideal state.

    1–2 hoursWith a few teammates

    Define · Problem statement

    Problem statement

    One or two sentences that say who has the problem, what it is, why it matters (to them and the business) and how you'll know it's solved. Includes the current and ideal state.

    • 1–2 hours
    • With a few teammates
    • Starter

    When to use it

    Before ideation, so everyone solves the same problem.

    How to run it

    1. Use the structure "[User] needs [need] because [insight]. We'll know it's solved when [metric].".
    2. Add the impact on users and on the business.
    3. Check it doesn't contain a solution.
    4. Agree it with the team.

    People: Product and design

    What you get

    An agreed problem to solve.

    Watch out

    A solution in disguise ("Users need a better dashboard").

    With AI

    AI can test a problem statement: ask whether it hides a solution, and for three alternative framings. Choose with the team.

    From practice
    At SecurEnvoy, the problem was framed as balancing security with usability for 2M+ enterprise users, which shaped every login decision.
  • Turning problems and insights into open questions that start ideation ("How might we help first-time buyers compare plans?").

    30–60 minutesGroup workshop

    Define · Problem statement

    How might we

    Turning problems and insights into open questions that start ideation ("How might we help first-time buyers compare plans?").

    • 30–60 minutes
    • Group workshop
    • Starter

    When to use it

    After research and before brainstorming.

    How to run it

    1. Take each insight.
    2. Rewrite it as "How might we…?".
    3. Make it neither too broad nor too narrow.
    4. Vote on the ones to work on.

    People: The team

    What you get

    A short list of questions to ideate on.

    Watch out

    Questions with the answer inside ("How might we add a chatbot?").

    With AI

    Ask AI to turn each insight into several How might we questions at different scopes, then pick the useful ones. It's fast at variety, not at judging which matter.

    From practice
    At J&R Marketing, turning client goals into user-centred questions was the start of 20+ product designs.

    Origin: Used at Procter & Gamble in the 1970s, where Min Basadur promoted problem-framing questions; later spread by IDEO and Google Ventures.

  • A tree from a desired outcome to the opportunities (user needs) that could move it, the solutions for each, and the experiments for each solution.

    1–2 hours to start, updated weeklyWith a few teammates

    Define · Problem statement

    Opportunity solution tree

    A tree from a desired outcome to the opportunities (user needs) that could move it, the solutions for each, and the experiments for each solution.

    • 1–2 hours to start, updated weekly
    • With a few teammates
    • Intermediate

    When to use it

    Continuous discovery, to keep ideas tied to outcomes.

    How to run it

    1. Start with one outcome.
    2. Map opportunities from research.
    3. Add 2–3 solutions per opportunity.
    4. Add a small experiment for each.

    People: Product trio

    What you get

    A visible link from every idea to the goal.

    Watch out

    Starting from favourite solutions; work down from the outcome and real user needs.

    With AI

    AI can suggest more solutions per opportunity and a small experiment for each. Keep the outcome and the opportunities grounded in your own research.

    From practice
    At Educative, hypothesis-driven experimentation kept design ideas tied to outcomes such as a 20% rise in course completion.

    Origin: Teresa Torres (Continuous Discovery Habits, 2021).

Ideate

Generate many ideas with the team, then choose the ones that help users and the business and can be built. Decide how you will measure them.

Brainstorming

  • A session to generate many ideas quickly, judging later.

    45–90 minutesGroup workshop

    Ideate · Brainstorming

    Brainstorming

    A session to generate many ideas quickly, judging later.

    • 45–90 minutes
    • Group workshop
    • Starter

    When to use it

    You have a clear problem or How might we question.

    How to run it

    1. Share the question and the research.
    2. Write ideas alone first (silently), then share.
    3. Build on others' ideas.
    4. Group and vote.

    People: 4–8 people

    What you get

    A wide set of ideas, grouped.

    Watch out

    Talking first lets the loudest or most senior person set the direction; write alone, then share.

    With AI

    Use AI to warm up the session or to add ideas once the team has written its own. Ideas from AI first anchor everyone on them, so add them last.

    From practice
    At PaysWell, design-to-development workshops with engineering cut handoff friction by 25%; bringing engineers into idea sessions early is part of that.
  • Each person folds a sheet into eight panels and sketches eight ideas in eight minutes.

    15–30 minutesGroup workshop

    Ideate · Brainstorming

    Crazy 8s

    Each person folds a sheet into eight panels and sketches eight ideas in eight minutes.

    • 15–30 minutes
    • Group workshop
    • Starter

    When to use it

    You need many rough variations fast, or people are stuck on their first idea.

    How to run it

    1. Fold a sheet into eight.
    2. Set a timer for 8 minutes.
    3. One idea per minute.
    4. Share and vote.

    People: Any group

    What you get

    Dozens of rough ideas in minutes.

    Watch out

    Polishing sketches; speed is the point.

    With AI

    Keep Crazy 8s on paper: it's about people's own ideas. Afterwards, AI tools can turn the strongest sketches into quick screens to compare.

    From practice
    At Horizontal Digital, AI-assisted workflows cut prototyping cycles by 35%, but quick paper sketching still starts most ideas.

    Origin: Google Ventures design sprint (Sketch day).

  • Everyone gets a few dot stickers (or digital dots) to vote on the options they think are strongest.

    10–20 minutesGroup workshop

    Ideate · Brainstorming

    Dot voting

    Everyone gets a few dot stickers (or digital dots) to vote on the options they think are strongest.

    • 10–20 minutes
    • Group workshop
    • Starter

    When to use it

    Narrowing many ideas down quickly.

    How to run it

    1. Post all options.
    2. Give each person 3–5 dots.
    3. Vote silently at the same time.
    4. Discuss the top clusters.
    5. The decider decides.

    People: Any group

    What you get

    A short list and a visible picture of the group's view.

    Watch out

    Voting in turn (people copy); vote at the same time.

    With AI

    Digital whiteboards run the vote and tally it with AI summaries of the winning clusters. The decider still makes the final call.

    From practice
    On Saudi Energy's platform, decisions had to work for 300+ product teams, so visible, quick decision methods mattered.
  • Quick hand or digital drawings to explore layouts and ideas before detailed design.

    30 minutes to a few hoursOn your own

    Ideate · Brainstorming

    Sketching

    Quick hand or digital drawings to explore layouts and ideas before detailed design.

    • 30 minutes to a few hours
    • On your own
    • Starter

    When to use it

    Early in design, before wireframes.

    How to run it

    1. List what the screen must do.
    2. Sketch several layouts fast.
    3. Annotate.
    4. Share and pick the strongest parts.

    People: Any designer

    What you get

    Rough concepts to discuss and combine.

    Watch out

    Going digital and detailed too early; it narrows the options.

    With AI

    AI can turn a photo of a sketch into an editable wireframe or a first layout. Sketch first anyway; it keeps the thinking yours.

    From practice
    At Horizontal Digital, AI-assisted workflows with Figma Make, Cursor and Claude cut prototyping cycles by 35%, starting from rough concepts.
  • A five-day process to answer a critical question through design, prototyping and testing with users: Map, Sketch, Decide, Prototype, Test.

    5 daysGroup workshop

    Ideate · Brainstorming

    Design sprint

    A five-day process to answer a critical question through design, prototyping and testing with users: Map, Sketch, Decide, Prototype, Test.

    • 5 days
    • Group workshop
    • Advanced

    When to use it

    A big, risky decision with enough uncertainty to justify five focused days.

    How to run it

    1. Monday: map the problem and pick a target.
    2. Tuesday: sketch solutions, each person alone.
    3. Wednesday: decide and storyboard the prototype.
    4. Thursday: build a realistic prototype.
    5. Friday: test it with five users.

    People: 5–7 people including a decider

    What you get

    A tested prototype and evidence for the decision.

    Watch out

    Running one for a small question, or without the decider in the room.

    With AI

    AI shortens the Prototype day the most: tools like Figma Make can build a realistic prototype from the storyboard in hours. Keep Map, Sketch and Decide with people.

    From practice
    At Tkxel, the Nitro League experience went from idea to a $5M raise in 20 days, the pace a sprint is built for.

    Origin: Jake Knapp, John Zeratsky and Braden Kowitz, Google Ventures (Sprint, 2016).

Prioritisation

  • A 2×2 placing ideas by user or business impact and the effort to build them.

    30–60 minutesWith a few teammates

    Ideate · Prioritisation

    Impact and effort matrix

    A 2×2 placing ideas by user or business impact and the effort to build them.

    • 30–60 minutes
    • With a few teammates
    • Starter

    When to use it

    Choosing what to do first from a list of ideas.

    How to run it

    1. Estimate impact with product and design.
    2. Estimate effort with engineering.
    3. Place each idea.
    4. Start with high impact, low effort.

    People: Product, design, engineering

    What you get

    A first order of work.

    Watch out

    Guessing effort without engineers.

    With AI

    Ask AI for a first impact and effort estimate per idea with its reasons, then correct it with engineering. Effort guesses without engineers are the usual failure, AI or not.

    From practice
    At Tkxel, design governance across 30+ concurrent projects meant deciding quickly what mattered most.
  • Sorting features by how they affect satisfaction: must-be (expected; missing ones cause frustration), performance (more is better), attractive (delight when present, no complaint when absent), indifferent and reverse.

    1–2 weeksOn your own

    Ideate · Prioritisation

    Kano model

    Sorting features by how they affect satisfaction: must-be (expected; missing ones cause frustration), performance (more is better), attractive (delight when present, no complaint when absent), indifferent and reverse.

    • 1–2 weeks
    • On your own
    • Intermediate

    When to use it

    Deciding which features meet basic expectations and which differentiate.

    How to run it

    1. List features.
    2. Ask users two questions per feature (how they'd feel with it, and without it).
    3. Classify with the Kano table.
    4. Cover must-bes first, then choose performance and attractive features.

    People: 1 researcher, 20+ survey participants

    What you get

    Features grouped by their effect on satisfaction.

    Watch out

    Categories change over time; today's delight is tomorrow's must-be.

    With AI

    AI can generate the paired Kano questions for your feature list and classify the answers. Ask real users; AI can't tell you what delights them.

    From practice
    On Virtua, discovery and checkout features were balanced to reach 150K+ registered users in six months.

    Origin: Noriaki Kano and colleagues (1984).

  • Scoring each idea by Reach × Impact × Confidence ÷ Effort.

    1–2 hoursWith a few teammates

    Ideate · Prioritisation

    RICE scoring

    Scoring each idea by Reach × Impact × Confidence ÷ Effort.

    • 1–2 hours
    • With a few teammates
    • Starter

    When to use it

    Comparing many ideas with a consistent, discussable score.

    How to run it

    1. Estimate reach (people per period).
    2. Impact (for example 0.25–.
    3. Confidence (%).
    4. Effort (person-months).
    5. Sort by score and discuss outliers.

    People: Product with design and engineering

    What you get

    A ranked list with the reasons visible.

    Watch out

    False precision; the score starts the discussion, it doesn't end it.

    With AI

    AI can pull reach from analytics exports and keep the scoring sheet consistent. Confidence scores are a team judgement; don't let AI fill them.

    From practice
    At Educative, data from Mixpanel and Amplitude gave reach and impact estimates real numbers.

    Origin: Intercom (Sean McBride).

  • Sorting requirements into Must have, Should have, Could have and Won't have (this time).

    1–2 hoursGroup workshop

    Ideate · Prioritisation

    MoSCoW

    Sorting requirements into Must have, Should have, Could have and Won't have (this time).

    • 1–2 hours
    • Group workshop
    • Starter

    When to use it

    Agreeing the scope of a release with stakeholders.

    How to run it

    1. List requirements.
    2. Agree what "must" means (the release fails without it).
    3. Sort together.
    4. Record the won'ts.

    People: Product, design, stakeholders

    What you get

    An agreed scope with clear trade-offs.

    Watch out

    Everything becoming a must.

    With AI

    Ask AI to sort a requirements list into a first MoSCoW draft with reasons, then debate the musts with stakeholders.

    From practice
    flydubai's five websites were re-platformed to Sitecore XM Cloud in stages, so each release needed an agreed scope.

    Origin: Dai Clegg, Oracle (1994); used in DSDM.

  • A 2×2 of products by market growth and market share: stars, cash cows, question marks and dogs.

    1–2 hoursWith a few teammates

    Ideate · Prioritisation

    BCG matrix

    A 2×2 of products by market growth and market share: stars, cash cows, question marks and dogs.

    • 1–2 hours
    • With a few teammates
    • Intermediate

    When to use it

    Product-portfolio decisions, such as which products to invest in or retire.

    How to run it

    1. List products.
    2. Place each by growth and share.
    3. Decide invest, keep, test or retire for each.

    People: Product and business leadership

    What you get

    A portfolio view to guide investment.

    Watch out

    It judges business position, not user value; pair it with research.

    With AI

    AI can place products on the matrix from market data you give it and explain each position. Check the data; the matrix is only as good as the numbers.

    From practice
    At J&R Marketing, a portfolio of 20+ digital products across sectors needed decisions on where to invest design effort.

    Origin: Bruce Henderson, Boston Consulting Group (1970).

  • Comparing the expected benefit of a design change (revenue, savings, fewer support calls) with its cost.

    1–2 daysWith a few teammates

    Ideate · Prioritisation

    ROI estimate

    Comparing the expected benefit of a design change (revenue, savings, fewer support calls) with its cost.

    • 1–2 days
    • With a few teammates
    • Intermediate

    When to use it

    Making the case for design work to leadership.

    How to run it

    1. Pick one metric the change affects.
    2. Estimate the improvement from research or tests.
    3. Turn it into money or time.
    4. Compare with the cost.
    5. State the assumptions.

    People: Design with product and finance

    What you get

    A business case with clear assumptions.

    Watch out

    Hiding the assumptions; show the range, not one number.

    With AI

    AI is useful for building the model: formulas, ranges and a clear list of assumptions. Use your own numbers and say where each comes from.

    From practice
    At flydubai, the design system's results were reported as business numbers: 62% faster page assembly, 3.4x component reuse, 68% fewer design QA defects.

Execute

Design and communicate the solution: structure, flows, screens and content, built on shared principles and a design system, and handed to engineering with every case covered.

Framework

  • Organising and labelling content and features so people can find them: structure, navigation and naming.

    1–4 weeksOn your own

    Execute · Framework

    Information architecture

    Organising and labelling content and features so people can find them: structure, navigation and naming.

    • 1–4 weeks
    • On your own
    • Intermediate

    When to use it

    New products, growing products, or navigation people get lost in.

    How to run it

    1. Inventory the content.
    2. Run card sorting.
    3. Draft a sitemap.
    4. Test it with tree testing.
    5. Design navigation from it.

    People: 1 designer

    What you get

    A sitemap, navigation and labels tested with users.

    Watch out

    Copying the company's org chart into the navigation.

    With AI

    AI can inventory content, propose a first sitemap and suggest labels. Test it with card sorting and tree testing; AI labels tend to follow the system, not the users.

    From practice
    For the Pakistan Ministry of Defence ERP and MES platform, the information architecture covered procurement, inventory, production planning and quality control.
  • Diagrams of the steps through a task, including the happy path, errors, empty states, permissions and limits.

    1–3 days per flowWith a few teammates

    Execute · Framework

    User flows and edge cases

    Diagrams of the steps through a task, including the happy path, errors, empty states, permissions and limits.

    • 1–3 days per flow
    • With a few teammates
    • Intermediate

    When to use it

    Before detailed design, and before handoff.

    How to run it

    1. Draw the happy path.
    2. Add every decision and error (no connection, no results, no permission, timeouts).
    3. Add empty, loading and limit states.
    4. Review with engineering.

    People: Designer with engineering

    What you get

    A complete flow engineering can build from.

    Watch out

    Designing only the happy path; most support calls come from the rest.

    With AI

    Ask AI to list the edge cases of a flow: errors, empty states, permissions, timeouts. It's thorough at this; then design each case.

    From practice
    At SecurEnvoy, login flows for multi-factor authentication had to handle every failure case; authentication succeeded 99.7% of the time.
  • Low-detail layouts that show structure, content and behaviour without visual styling.

    Days per featureOn your own

    Execute · Framework

    Wireframes

    Low-detail layouts that show structure, content and behaviour without visual styling.

    • Days per feature
    • On your own
    • Starter

    When to use it

    Agreeing layout and priorities before visual design.

    How to run it

    1. Start from flows and content.
    2. Lay out each screen in greyscale.
    3. Use real content where possible.
    4. Review and iterate.

    People: 1 designer

    What you get

    Agreed screen structures.

    Watch out

    Lorem ipsum; real content changes the layout.

    With AI

    AI tools generate wireframes from a prompt or a sketch in minutes. Use them to explore options, then refine with real content and your design system.

    From practice
    At J&R Marketing, wireframing was part of a full design lifecycle that raised user engagement by 40%.
  • Interactive versions of a design, from clickable wireframes to near-real builds, used to test and explain.

    Hours to daysOn your own

    Execute · Framework

    Prototypes

    Interactive versions of a design, from clickable wireframes to near-real builds, used to test and explain.

    • Hours to days
    • On your own
    • Starter

    When to use it

    Testing a flow with users, or showing stakeholders how it will work.

    How to run it

    1. Decide the question the prototype must answer.
    2. Choose the fidelity that answers it.
    3. Build only the paths needed.
    4. Test and iterate.

    People: 1 designer

    What you get

    A prototype to test or present.

    Watch out

    Over-building; a prototype is for learning, not for shipping.

    With AI

    AI prototyping tools (Figma Make, Lovable and similar) build working prototypes from a description. Decide the question first, and check the prototype's accessibility before testing.

    From practice
    At Horizontal Digital, AI-assisted workflows (Figma Make, Cursor and Claude) cut prototyping cycles by 35%.

Guidelines

  • Shared foundations (tokens), components, patterns and guidelines that teams use to build consistent products.

    Months, then ongoingWith a few teammates

    Execute · Guidelines

    Design system

    Shared foundations (tokens), components, patterns and guidelines that teams use to build consistent products.

    • Months, then ongoing
    • With a few teammates
    • Advanced

    When to use it

    More than one team or product, or repeated design and build effort.

    How to run it

    1. Audit what exists.
    2. Define foundations and tokens.
    3. Build and document the most used components.
    4. Set governance (who decides, how changes are made).
    5. Measure adoption.

    People: A dedicated team

    What you get

    A design system teams adopt.

    Watch out

    Building components nobody asked for; start from what teams use most.

    With AI

    AI helps with documentation, token naming, finding inconsistent components and generating code from components. Governance and adoption remain people's work.

    From practice
    For flydubai's five websites on Sitecore XM Cloud: a 4-tier token architecture, a component QA gate and RACI-based governance (62% faster page assembly, 3.4x component reuse). For Saudi Energy: a system for 300+ product teams that cut UI inconsistencies by 30%.
  • Checking designs and builds against WCAG 2.2: contrast, keyboard use, screen reader output, focus, target size, content and more.

    1–5 days per product areaOn your own

    Execute · Guidelines

    Accessibility review

    Checking designs and builds against WCAG 2.2: contrast, keyboard use, screen reader output, focus, target size, content and more.

    • 1–5 days per product area
    • On your own
    • Intermediate

    When to use it

    In every phase: design reviews, before handoff, before release.

    How to run it

    1. Check designs (contrast, focus order, labels, target size).
    2. Run an automated scan.
    3. Test by hand with a keyboard and a screen reader.
    4. Record results per criterion and fix by role (design, code, content).

    People: Designer or accessibility specialist

    What you get

    A report of issues by WCAG criterion and owner.

    Watch out

    Trusting an automated scan alone; it finds only part of the issues.

    With AI

    AI can explain failures, suggest fixes and draft alt text. It can't judge whether a page works with a screen reader or keyboard; test those by hand.

    From practice
    Accessibility-first practice on WCAG 2.2 reached 95%+ compliance at Horizontal Digital, and 88% WCAG 2.2 / DGA compliance on Saudi Arabia's national energy platform (35M+ users).

Content

  • Writing the words in the interface (labels, buttons, errors, help, empty states) so they are clear, short and consistent.

    OngoingOn your own

    Execute · Content

    UX writing

    Writing the words in the interface (labels, buttons, errors, help, empty states) so they are clear, short and consistent.

    • Ongoing
    • On your own
    • Starter

    When to use it

    From wireframes on; never left to the end.

    How to run it

    1. Use the users' words from research.
    2. Write the action on buttons ("Save changes", not "OK").
    3. Write errors that say what happened and how to fix it.
    4. Keep a word list.

    People: Designer or content designer

    What you get

    Interface copy and a terminology list.

    Watch out

    Internal jargon; users don't know the team's names for things.

    With AI

    AI is fast at drafting labels, error messages and variants in a set tone. Give it your word list, and review for accuracy: confident wrong instructions are worse than none.

    From practice
    At SecurEnvoy, streamlining login experiences cut login-related support tickets by 30%.

Handoff

  • Giving engineering everything needed to build: specs, states, flows, tokens, assets, accessibility notes and acceptance criteria, and walking through it together.

    Hours to days per featureWith a few teammates

    Execute · Handoff

    Design handoff

    Giving engineering everything needed to build: specs, states, flows, tokens, assets, accessibility notes and acceptance criteria, and walking through it together.

    • Hours to days per feature
    • With a few teammates
    • Intermediate

    When to use it

    Before build, and continuously during it.

    How to run it

    1. Check every state and edge case is designed.
    2. Link components to the design system.
    3. Annotate behaviour and accessibility (focus order, labels).
    4. Hold a walkthrough.
    5. Review the build together.

    People: Designer with engineering

    What you get

    A build-ready design and a shared understanding.

    Watch out

    Handing off files without talking; questions surface late and get guessed.

    With AI

    AI can draft specs and acceptance criteria from designs and turn components into code. Walk through it with engineers anyway; the conversation catches what specs miss.

    From practice
    At PaysWell, design-to-development workshops cut handoff friction by 25% and raised sprint velocity by 20%. At Horizontal Digital, automated handoff improved design-to-development efficiency by 40%.
  • A structured review of work in progress against the goals, with specific, kind feedback.

    30–60 minutes a weekWith a few teammates

    Execute · Handoff

    Design critique

    A structured review of work in progress against the goals, with specific, kind feedback.

    • 30–60 minutes a week
    • With a few teammates
    • Starter

    When to use it

    Regularly during design, before decisions are fixed.

    How to run it

    1. The designer states the goal and the feedback wanted.
    2. Show options, not one answer.
    3. Reviewers ask questions and relate feedback to goals.
    4. The designer decides what to act on.

    People: Design team, sometimes product and engineering

    What you get

    Better work and a team that learns together.

    Watch out

    Feedback on taste ("I don't like blue") instead of goals.

    With AI

    Ask AI for a heuristic review of a screen before the critique, to clear the obvious issues. Keep the critique itself between people, focused on the goals.

    From practice
    At Horizontal Digital, regular design critiques were part of building and mentoring a design team of 12+.

Principles to design by

Not methods to run, but principles to check every design against.

Usability heuristics
Jakob Nielsen's 10 heuristics (1994, wording updated 2020), such as visibility of system status and error prevention. The basis of a heuristic evaluation.
Don't make me think
Steve Krug (2000): pages should be self-evident; people scan and muddle through rather than read.
Consistency and standards
Follow the platform's conventions and your own design system, so people reuse what they already know.
Gestalt principles
Proximity, similarity, continuity, closure and common region: how people group what they see.
Accessibility first
WCAG 2.2 at level AA as the baseline, from the first sketch, not as a final check. Open the WCAG 2.2 report tool (opens in a new tab)
Platform standards
Native patterns for iOS, Android and the web (navigation, controls, gestures) before custom ones.

Validate

Check the solution with real people and real data. Look for what you expected and what you didn't, then decide what to improve, what to test next and what new opportunities appear.

Quantitative

  • Showing two (or more) versions to random groups of users and comparing a metric.

    1–4 weeksWith a few teammates

    Validate · Quantitative

    A/B testing

    Showing two (or more) versions to random groups of users and comparing a metric.

    • 1–4 weeks
    • With a few teammates
    • Intermediate

    When to use it

    You have enough traffic and one clear metric, and want proof one version performs better.

    How to run it

    1. Write a hypothesis and the metric.
    2. Work out the sample size and duration before starting.
    3. Split traffic randomly.
    4. Run to the planned end.
    5. Check significance, then look at segments.

    People: Designer, engineer, analyst

    What you get

    Evidence of which version performs better on the metric.

    Watch out

    Stopping early when the result looks good ("peeking"), or testing with too little traffic.

    With AI

    AI helps write variants, calculate sample sizes and read results. Fix the hypothesis and the stopping point before the test starts, whoever analyses it.

    From practice
    At GreatPeople, A/B tests, heatmaps and feedback loops improved task completion by 25% and cut support tickets by 20%.
  • Tracking the agreed success metrics after release, with a baseline and a review rhythm.

    OngoingWith a few teammates

    Validate · Quantitative

    Metrics monitoring

    Tracking the agreed success metrics after release, with a baseline and a review rhythm.

    • Ongoing
    • With a few teammates
    • Starter

    When to use it

    After every release that was meant to change behaviour.

    How to run it

    1. Record the baseline before release.
    2. Set up a dashboard with the HEART or OKR metrics.
    3. Review at fixed points (1 week, 1 month).
    4. Compare with the goal and decide.

    People: Product, design, data

    What you get

    Whether the change worked, and what to do next.

    Watch out

    Changing several things at once so you can't tell which one worked.

    With AI

    AI alerts in analytics tools flag unusual changes after a release. Check alerts against what changed; seasonality and campaigns cause false alarms.

    From practice
    Saudi Energy's platform reached 85% adoption, 42% fewer support calls and 89% customer satisfaction: numbers that only count when they're tracked after release.
  • Releasing to a small group (a region, a percentage, a customer) before everyone.

    1–4 weeksWith a few teammates

    Validate · Quantitative

    Pilot release

    Releasing to a small group (a region, a percentage, a customer) before everyone.

    • 1–4 weeks
    • With a few teammates
    • Intermediate

    When to use it

    Risky changes, or when you need real-use data before the full launch.

    How to run it

    1. Choose the group.
    2. Define success and stop criteria.
    3. Release and monitor.
    4. Collect feedback.
    5. Decide to expand, change or roll back.

    People: Product, engineering, design

    What you get

    Real-world evidence with limited risk.

    Watch out

    A pilot group unlike your real users.

    With AI

    AI can watch pilot feedback and metrics and summarise issues daily. Keep the go or no-go decision with the team, against the criteria you set.

    From practice
    At GreatPeople, continuous optimisation with A/B tests and feedback loops served 50K+ daily users.
  • The System Usability Scale: 10 statements rated 1–5, giving a score from 0 to 100.

    Minutes per participantOn your own

    Validate · Quantitative

    SUS questionnaire

    The System Usability Scale: 10 statements rated 1–5, giving a score from 0 to 100.

    • Minutes per participant
    • On your own
    • Starter

    When to use it

    After usability tests or releases, to compare perceived usability over time or with other products.

    How to run it

    1. Give the standard 10 items after the tasks.
    2. Score with the standard formula.
    3. Compare with your previous score; the average across many studies is about 68.

    People: 1 researcher

    What you get

    One comparable usability score.

    Watch out

    Reading 68 as 68%; it's a score, not a percentage. Don't change the wording.

    With AI

    AI can score SUS answers and chart them over time. Don't let it reword the 10 statements; the score is only comparable with the standard wording.

    From practice
    At GreatPeople, manager satisfaction scores improved by 45% after the redesign, the kind of change a standard score makes visible.

    Origin: John Brooke (1986); the average of 68 is from Jeff Sauro's analysis (MeasuringU).

Qualitative

  • Watching people try to complete real tasks with a design or product while thinking aloud.

    1–2 weeks per roundOn your own

    Validate · Qualitative

    Usability testing

    Watching people try to complete real tasks with a design or product while thinking aloud.

    • 1–2 weeks per round
    • On your own
    • Starter

    When to use it

    At every stage, from paper prototype to live product.

    How to run it

    1. Write 3–5 realistic tasks.
    2. Run a pilot session to fix the script.
    3. Test with about 5 people per user group.
    4. Observe without helping.
    5. List issues by severity and fix, then test again.

    People: 1 facilitator, 1 note-taker, about 5 participants per group

    What you get

    The problems people actually meet, by severity.

    Watch out

    Leading the participant ("Click the blue button"), or testing once and stopping; several small rounds find more.

    With AI

    AI tools transcribe sessions, mark moments of hesitation and draft the findings. Watch the sessions yourself; AI summaries miss body language and context.

    From practice
    At SellAnyCar, guerrilla and remote usability testing with 100+ users improved conversion from views to inquiries by 30%.
  • Experts review an interface against usability heuristics (usually Nielsen's 10).

    1–3 daysWith a few teammates

    Validate · Qualitative

    Heuristic evaluation

    Experts review an interface against usability heuristics (usually Nielsen's 10).

    • 1–3 days
    • With a few teammates
    • Intermediate

    When to use it

    Quickly finding obvious problems before testing with users, or when users are hard to reach.

    How to run it

    1. 3–5 evaluators review alone.
    2. Each notes issues with the heuristic and severity.
    3. Combine and remove duplicates.
    4. Prioritise fixes.

    People: 3–5 evaluators

    What you get

    A list of likely problems, ranked.

    Watch out

    One evaluator finds only about a third of the problems; use several, and still test with users.

    With AI

    AI can run a first heuristic pass on screenshots and list likely issues. Use it as one extra evaluator, not as the panel; it misses context and flow problems.

    From practice
    At SecurEnvoy, a heuristic evaluation framework streamlined login experiences while keeping security compliance.

    Origin: Jakob Nielsen and Rolf Molich (1990).

  • The team steps through a task as a new user, asking at each step whether the user would know what to do and understand the result.

    1–3 hours per taskWith a few teammates

    Validate · Qualitative

    Cognitive walkthrough

    The team steps through a task as a new user, asking at each step whether the user would know what to do and understand the result.

    • 1–3 hours per task
    • With a few teammates
    • Intermediate

    When to use it

    Checking how learnable a flow is, especially for first-time use.

    How to run it

    1. Define the user and the task.
    2. At each step ask: will they try this, will they see how, will they understand the feedback.
    3. Note failures.
    4. Fix and repeat.

    People: 2–4 team members

    What you get

    The steps where new users get stuck.

    Watch out

    Walking through as an expert; you know too much.

    With AI

    Ask AI to play a first-time user at each step and say what it would expect. It's a cheap extra perspective, not a replacement for testing with new users.

    From practice
    On Nitro League, a streamlined onboarding made wallet connection and asset management simpler for first-time players.
  • Participants find items in a text-only version of your navigation tree.

    1–2 weeksOn your own

    Validate · Qualitative

    Tree testing

    Participants find items in a text-only version of your navigation tree.

    • 1–2 weeks
    • On your own
    • Intermediate

    When to use it

    Checking an information architecture before designing screens.

    How to run it

    1. Build the tree from your sitemap.
    2. Write 8–10 find-it tasks.
    3. Run it online.
    4. Look at success, directness and where people went first.

    People: 1 designer, 50+ participants for reliable rates

    What you get

    Which parts of the navigation work and which mislead.

    Watch out

    Tasks that use the same words as the labels.

    With AI

    AI can draft find-it tasks from your top user goals and analyse where people went wrong. Write tasks in users' words, not the menu's.

    From practice
    The Pakistan Ministry of Defence ERP had an information architecture across four operational areas, in a project that cut procurement cycle times by 60%.
  • Two quick online tests. First-click: where people click first to complete a task, a strong predictor of success. Five-second: what people remember after seeing a page for five seconds.

    1–3 daysOn your own

    Validate · Qualitative

    First-click and five-second tests

    Two quick online tests. First-click: where people click first to complete a task, a strong predictor of success. Five-second: what people remember after seeing a page for five seconds.

    • 1–3 days
    • On your own
    • Starter

    When to use it

    Checking whether a page's layout and message are clear at a glance.

    How to run it

    1. Show a screen and one task or question.
    2. Record the click or the recall.
    3. Compare with what you intended.

    People: 1 designer, 20–50 participants online

    What you get

    Whether people see what matters, and where they'd go.

    Watch out

    Reading too much into one screen; use it alongside usability testing.

    With AI

    AI attention heatmaps predict where people may look; treat them as a hint and confirm with a real first-click test.

    From practice
    On Virtua, visual discovery was central to the marketplace, and 92% of first-time buyers completed their task.
  • Measuring where people look on a screen with specialist hardware or software.

    2–4 weeksOn your own

    Validate · Qualitative

    Eye tracking

    Measuring where people look on a screen with specialist hardware or software.

    • 2–4 weeks
    • On your own
    • Advanced

    When to use it

    Questions about attention: are people seeing a warning, an offer, a label?

    How to run it

    1. Define what you want to learn.
    2. Set up and calibrate.
    3. Run realistic tasks.
    4. Analyse gaze plots and heatmaps.

    People: Specialist, lab or tool

    What you get

    Where attention goes and what gets missed.

    Watch out

    Heatmaps need many participants to be reliable (NN/g advise about 30 or more); for most questions a usability test is cheaper.

    With AI

    AI-predicted attention maps are a cheap first look. For decisions, use real eye tracking or a usability test; predictions are trained on averages.

    From practice
    At GreatPeople, heatmap analysis helped improve task completion by 25%.
  • Collecting and grouping feedback from support tickets, reviews, surveys and in-product feedback, to find recurring problems.

    OngoingOn your own

    Validate · Qualitative

    User feedback analysis

    Collecting and grouping feedback from support tickets, reviews, surveys and in-product feedback, to find recurring problems.

    • Ongoing
    • On your own
    • Starter

    When to use it

    Continuously, and before planning the next round of work.

    How to run it

    1. Gather feedback in one place.
    2. Tag by topic and severity.
    3. Count and track themes over time.
    4. Link themes to metrics and research.

    People: Design with support

    What you get

    The top recurring problems, with evidence.

    Watch out

    Only hearing the loudest users; feedback is a signal, not a sample.

    With AI

    AI is good at tagging large volumes of tickets and reviews by theme and sentiment. Read a sample of each theme to make sure the label is right.

    From practice
    On Saudi Energy's platform, support calls fell by 42%; support contacts are one of the richest sources of feedback.

Need a hand?

Make the process your team's own.

A list of methods doesn't make a process. I help teams choose the few methods that fit their product and pace, run them together the first time, and set up the rituals, templates and handoffs that keep them going.

Design process and DesignOps

  • A review of how your team works today, from research to release
  • A design process fitted to your product, team and timelines
  • The first round run with your team: research, workshop or test
  • Templates, critiques and handoff rituals that make it stick

4 → 12designers at Educative, with design operations that cut review cycles by 40%