UX Research

UX research aids decision-making by giving data-backed answers. Without it, teams build products based on guesses and assumptions. But data also needs analysis to make actionable decisions.

Research continues throughout the product lifecycle:

• Early: find out user needs & goals and how your product can help.
• Mid-stage: test assumptions.
• Later: eavluate usability & user perceptions.
• Post-launch: find out what’s wrong to keep users engaged.

UX research plan:

• Set goals & questions
• Identify available resources: budget & timeframes
• Choose methods & participants
• Set timeline
• Collect & analyze data: the format of results include reports, presentation, recorded sessions

Empathy: Think about “what users experience and why it matters to them” instead of “what users do”.

Lean research methods: get quick feedback from users to build MVPs and test right away

• Guerrilla testing: get quick feedback from people in public spaces, 5-10mins/session
• Remote unmoderated testing: users complete tasks using recording tools
• Rapid surveys

Research for an existing product:

• Churn & retention: data like drop off & decrease usage; and interviews
• Optimize: session recordings, heatmaps
• Validate new features & improvements: A/B testing to see conversion & engagement, prototype testing to see user interest

Usability heuristics

Visibility of system status: Inform users about what’s happening and provide relevant & timely feedback for user actions. Users know where they are, what they can do next, if their actions are processed. E.g. change of color, animated progress tracker, sound, haptic, notification, modal dialog, inline message.

Match between system and real world: Use language, interactions, and concepts familiar to users. Simple language, universal icons, interactions mimic reality like flipping pages or pressing buttons, elements placed where users naturally expect them.

User control and freedom: allow people to undo, correct, and reset. Undo/Redo, Back, Cancel, Close.

Consistency and standards: The same words & actions throughout a product.

Error prevention: Suggestions while typing, error notifications, smart defaults, forgiving formatting & adjust data automatically.

Recognition rather than recall: Show options that users have seen before rather than requiring them to remember details. Search history and recent items, autocomplete and suggestions, previews (emails, file thumbnails), saved forms info.

Flexibility and efficiency of use: Provide tailored experience for novice & experienced users. Users can approach tasks in multiple ways, customize interface, use accelerators like keyboard shortcuts and macros.

Aesthetic and minimalist: People have more tolerance for minor usability issues in interfaces that look good.

Error recognition, diagnosis, and recovery: Error message should be in plain, human language instead of technical terms or codes; doesn’t sound judgemental; is clear what’s wrong; how to resolve.

Help and documentation: Proactive help like onboarding flows and tips + Reactive help like tutorials, videos, documentation.

Ethics

3 pillars of ethical research:

• Respect user privacy
• Obtain informed consent
• Minimize bias

Participants may feel nervous or withhold information. To create a safe environment for them, reassure them by explaining the session format, letting them skip questions, and being clear that there are no right or wrong answers.

For sensitive topics such as health, financial or personal difficulties, conduct one-on-one or small-group sessions.

Biases

Confirmation bias happens when you focus on evidence that supports existing assumptions and overlook feedback that contradicts them.

• Ask open-ended questions so participants can use their own words
• Instead of “Was this easy to use?”, consider “How did you feel about the language use?”
• Active listening instead of filtering responses to what you want to hear
• Include diverse participants so you get varied feedback

False consensus bias happens when you assume most people think, feel, or behave the way you do. Treat your assumptions as hypotheses, not facts, and test them with users.

Recency bias happens when you focus more on what participants say or do toward the end of the session than at the beginning. People often remember recent moments more vividly than the earlier ones, even when they can be equally important. Users may struggle for the first 5mins and breeze through the task during the last min, but you focus on that smooth finish.

• Take structured, detailed notes throughout the session
• Record sessions (with consent)
• Review notes immediately after each session
• 1 researcher ask question and 1 researcher take notes

Primacy bias happens when you focus on the first participant than later participants. What the first one say or do may distort what you hear from others after.

Implicit bias refers to attitudes or stereotypes we hold about certain groups of people without being aware of them. E.g. you assume older people have hearing or cognitive difficulties, so you speak more slowly and loudly and make them feel disrespected.

Sunk cost fallacy is the tendency to keep investing in something because you’ve already spent too much on it, even when stopping is the smarter choice.

Research findings can be presented with bias. If you cherry-pick quotes that support a preferred direction or soften findings, you’re letting bias shape the output. A single quote from one participant is an individual opinion. If it’s said by 8/10 participants, it’s a pattern that’s worth acting on.

UX research methods

Qualitative

The why instead of the numbers.

When to use: smaller participant groups & help point direction rather than confirm a scale.

Ethnographic research

Immerse yourself in users’ natural environments to observe how they actually behave, as there may be distortion that comes from users knowing they’re being watched. For example, you may see servers input orders while carrying plates, and this doesn’t pop up in an interview.

Data collection: Field notes, photos, videos, artifact analysis.

Researchers may observe or participate alongside users.

Cons: expensive & time-consuming.

Use when: assumptions about users are weak or untested, wrong design direction brings real consequences.

Contextual inquiry

Observation + in-context interview. Can be in person or remotely with screen sharing. A session typically runs 2 hours & has flexible structure.

4 principles:

• Context: conduct where user naturally works
• Partnership: treat user as expert & researcher as learner
• Interpretation: share observations with user & ask them to confirm or correct
• Focus: on your research goals

Workarounds are signs your product need to improve to meet user needs.

Use when: for understanding complex workflows & uncovering behaviors users can’t easily describe in a standard interview.

Diary studies

Participants self-report their behaviors, thoughts, and feelings over an extended period. Users log entries at regular intervals. Then, researchers follow up with interviews. For example, users make repeat purchases from brands get a diary kit with questions about relationships, routines, expectations.

Use when: for understanding behaviors that unfold gradually; for understanding how users perform specific tasks in their time and context; discovery phase in design progress.

Logging entries should be easy, low-friction for participants.

User interviews

• Structured: fixed set of questions, fixed order → consistent, comparable responses
• Unstructured: useful in early discovery
• Semi-structured: there are key questions but flexible for follow up questions

Use when: gather feedback, build personas, understand user needs & goals, explore attitudes

Usability testing

Participants perform tasks while researchers observe and listen to feedback.

Formats:

• Moderated: a facilitator guides participant through tasks in real time
• Unmoderated: participants complete tasks on their own using a testing platform
• Guerilla testing

Focus groups

A facilitator leads a group of 6-9 participants through a structured discussion about their experience with a product/service. A session typically runs 1-2 hours.

Focus groups are most useful during early discovery. They collect what people say, how they frame problems, and what language they use. They help understand users’ mental models, vocabulary, and general attitudes.

Cons: may carry bias risk as dominant voices can shape the conversation and others don’t share their genuine view. Facilitators can use techniques like diverge-converge, where participants write their responses before sharing.

Quantitative

Collect numerical data about how users interact with a product. Easy to compare, track, and present to stakeholders. Answer questions like how many, how often, and how long.

A/B testing

• Choose a variable to test
• Define a single goal and form a hypothesis around it
• Split audience equally & randomly

Multivariate testing: test several design elements at once, capture how combinations of variables interact.

Eye tracking

It captures:

• Fixations: where the gaze rests
• Saccades: rapid movements between fixations
• Overall scan path: which elements attract attention & which go unnoticed

Output:

• Heatmap: needs large sample to be reliable (30-40 participants)
• Session replays

Retrospective think-aloud, where participants comment while watching their session replay, yields more natural insights.

Surveys

• Define goal
• Logically structure questions
• Write questions in plain language
• Include open-ended questions
• Keep it short

Surveys shouldn’t be used to replace usability studies and interviews. Point scales to ask how likely users will use a product or recommend it don’t reflect reality.

System Usability Scale (SUS) scores

10 statements + 5 response options ranging from Strongly Agree to Strongly Disagree.

68: above average <68: signals usability problems

Web analytics

Useful reports:

• Pageview: track where users enter & exit a product, paths, devices
• Behavior flow: spot drop off or unexpected routes
• Goals & funnels: define target actions & track how many users reach each step
• Event tracking: measure how users interact with specific elements
• Time on task: surface friction points

Usability benchmarking

Participants complete realistic tasks to produce metrics like task completion rates, time on task, error rates. Can be conducted in person or remotely, with or without a moderator. 30-40 participants.

Desirability studies

Measure aesthetic appeal & identify visual design directions that resonate with users and support the brand image. Participants are shown product images, prototypes, or versions of the same interface, then they select adjectives from a list to describe each design.

Choose a UX research method

Define research goal

Ask these questions:

• What do I want to learn? Why?
• Who your users actually are? Their needs & wants? How do they currently interact with the product? Can they use it? Do they enjoy using it?

Match research methods to design phases

• Early: field studies, ethnography, diary studies, interviews, surveys, participatory design, concept testing, focus groups.
• During design: qualitative usability testing, card sorting, tree tresting, usability testing, moderated/unmoderated remote testing, first click testing, task analysis, A/B testing, accessibility testing
• After release: usability benchmarking, unmoderated UX testing, A/B testing, analytics, surveys (NPS)

Qualitative vs quantitative research methods

Qualitative:

• Answer “why”: Why do users behave this way? Why does this part feel frustrating?
• To understand users in depth, to look for patterns rather than numbers
• Results can be influenced by poor questions, misunderstandings, researcher bias

Quantitative:

• Answer “What do users do?”, “How many people use this?”, “How often does this error occur?”
• To spot trends, test hypotheses, and compare options
• Can’t explain the reasoning

Attitudinal vs behavioral research methods

Attitudinal:

• Help understand users’ mental models & opinions: what they believe about a product, how they expect it to work, what they think about their experience
• Card sorting reveals how users mentally categorize info, which informs your information architecture; Surveys help track attitudes and opinions; Focus groups reveal reactions to concepts or designs.

Behavioral:

• Focus on what users actually do when they interact with a product/service
• A/B testing helps compare different designs; eye tracking shows how users navigate a design.

In practice, most research methods blend both types to a certain extent. For example, usability testing + field studies.

Choose methods by usage context

• Natural use: Users use the product on their own terms. Ethnographic field studies, intercept surveys, analytics
• Scripted use: Researcher guides participants. Benchmarking study
• Limited use: Participants use one function or aspect of the experience rather than the product as a whole. Participatory design, concept testing, desirability study
• Not using the product: Card sorting, concept testing and study

Methods for user needs research

Needs are often things users can’t easily articulate. If users struggle to explain their needs, it’s a signal to observe rather than interview.

• Contextual inquiry and ethnography
• Interviews
• Surveys & questionnaires
• Diary studies

Methods for user wants research

Wants are often tied to preferences, expectations, and desirability.

• A/B testing: versions of a design
• Focus groups: see how people think about your product
• Rapid prototyping: test whether a design direction resonates with users

Methods for usability research

Can users actually use what you’ve built?

• Usability testing
• Card sorting: run before you commit to an information structure
• Tree testing: validate a structure

Card sorting

Participants group labels into categories, which reveals mental models and helps teams build information architecture.

Open card sorting is useful when you want to learn how users organize info mentally, how they search for content, and build a new information architecture. Closed card sorting is useful when you want to assess your current labeling system, find out confusing or redundant categories, and prioritize user actions.

Prepare 30-60 cards and avoid cards with similar words or synonyms. Give participants 15-20 mins, with an extra 5 mins to sort. You acan process results using a spreadsheet matrix, or tools like Optimal Workshop or UserZoom.

Surveys

The order of the questions can affect results due to bias. If the earlier questions ask about satisfaction with a successful feature, participants are more likely to give positive responses for later questions.

In written and online surveys, participants tend to gravitate toward options at the top of the list (primacy effect). In verbal surveys, the last options stay freshest in participants’ memory and are more likely to be chosen (recency effect).

To reduce bias, randomize the order of the questions and options.

Start broad and narrow down as you go: Simple, quick, close-ended questions → Measurement, close-ended questions (rating scales, multiple choice, yes/no) → Open-ended questions. Too many and too early open-ended questions lead to fatigue and lower completion rates.

Each question should measure one thing.

Tell people upfront how long the survey takes and how many questions to expect. Add a progress indicator. Let them skip questions. Use conditional logic to show follow-up questions to relevant respondents.

Use AI to generate a first draft. Paste that draft back and ask it to flag leading, loaded, double-barreled questions, vary question types. You review. And ask AI to take the survey from a participant’s perspective to find out issues early.

Best practices for UX research questions

What’s the theme of your research? This will help you narrow down to actionable questions & filter off-topic answers.

Start with warm-up questions. They’re easy to answer & closely related to the broader topic. For example, if you’re studying how users make purchases on an e-commerce app, you can ask:

• How often do you shop on e-com website?
• Which e-com app do you use often?
• When was the last time you bought something through an app?

Because people tend to forget details and tell a neater version of their behavior, ask them about the last time they did something instead of how they generally behave. E.g. “What did you do the last time you came across an unhelpful 404 page?” instead of “How do you react to unhelpful error message?“.

Avoid leading questions that carry your own judgment. Instead of “how difficult”, “what about this is beautfiful”, use “how would you describe” or “how did you feel”.

User introduction question (lifestyle, habits, routines) tell you how they live, what they care about, or how their daily context might affect their relationship with your product.

Topic-specific questions:

• How do you currently [complete this task]?
• Walk me through the last time you [couldn’t complete this task]. What happened?
• What does your process look like when you [do this task]?
• How does that affect you when things go wrong?
• What would a better experience look like for you?

Use concept testing questions to find out whether your product resonates with target users:

• First impressions of this product?
• Your expectations?
• What makes you hesitant to use it?
• How does this compare to how you currently solve this problem?
• What would you need to change for this to feel useful to you?

Ask about a specific moment may yield better responses than general questions:

• What part of this product do you rely on most?
• Tell me about a time when the product didn’t do what you expected.
• What would you change if you could change one thing?

Participant recruitment for UX research

• Get informed consent before every research session

The numner of participants

• Most qualitative usability studies, 5 is a solid starting point.
• Quantitative: at least 20
• Card sorting: at least 15
• Eye tracking: 39

Participant critieria

Behavioral characteristics tend to matter more than demographic ones. 2 people of different ages who book travel at the same frequency > 2 people of the same age with opposite habits.

UX designers, marketers, IT specialists tend to analyze interfaces more, giving expert feedback rather than realistic user behavior.

Designing for the edge cases and people at the extremes of your user spectrum (power users - novices) tend to improve the experience for everyone in between.

Screening questions to find qualified participants

• Avoid letting candidates know the study purpose because they may adjust their answers to qualify.
• Start broad and get specific later.
• Ask about general digital habits before narrowing down.
• Add a follow-up phone call for sensitive or revealing questions.

Conducting user interviews

Your role is to hear and document what the participant shares, not to relate to it. Avoid empathetic phrases like “I understand how you feel” or “something similar happened to me.”

Analyze UX research

Conduct debrief sessions

A research debrief is a meeting held right after a research session, while memory is still fresh. A session typically runs 30mins-1hr, but shorter ones are also better than none.

Before a session starts, prepare a whiteboard or a digital workspace and prompts:

• What stood out
• Key quotes
• Surprises
• Open questions

The debrief organizer can set ground rules about the format of the session, define and explain terms with participants, and give time boundaries. Keep the focus on observations first, and if ideas for solutions come up early, reserve a space for them to come back later.

One person can take notes while everyone discusses, or give each person a short list of questions to fill out during the session. Once the debrief is done, the notes can be collected to share with team or stakeholders.

Use AI to generate summary after the session, using tools like Looppanel or Dovetail. AI can also surgace moments worth discussing, like emotional or contradictory moments. It can also spot patterns after 2-3 sessions.

UX research analysis

Raw data tells you what happened. Analysis tells you why.

Attitudinal data captures what users say, think, and feel. The most common method for analyzing attitudinal data is thematic analysis—assigning short labels (codes) to meaningful quotes or observations & looking for patterns & grouping into theme. The limitation is users don’t always do what they say they will.

Behavioral data captures what users actually do. The core approach to analyzing behavioral data is pattern recognition. And then coding patterns & grouping codes into themes. The challenge is it doesn’t tell you why, so follow up with an interview or a short survey to collect attitudinal data.

Map each key finding to a specific research goal before your readout. If a finding doesn’t connect to any goal, ask yourself whether it warrants a separate follow-up study.

When synthesizing research findings, aim for 3 to 8 insights per study. Fewer may suggest your research scope was too narrow. More often means you haven’t been critical enough about what genuinely matters.

Quantitative data analysis

Tools: SPSS, JMP, Stata, R

Methods:

• Cross-tabulation: compare feature usage across age groups or devices
• Max-diff analysis: (best-worst scaling) measure importance of features to know what to prioritize
• Conjoint analysis: It breaks a product down to attributes (price, speed, design) and levels (low, medium, high), and determines which combination resonates most with audience
• Gap analysis: measures the distance between where users expect your product to be and where it actually is (satisfaction scores)
• Trend analysis: tracks how a metric changes over time & why
• Sentiment analysis: uses natural language processing tools to process open-ended text responses & categorize feedback to surface patterns

Qualitative data analysis

Methods:

• Thematic analysis: coding & grouping codes into themes
• Content analysis: count how often certain words or topics appear to know how often a concern comes up
• Narrative analysis
• Affinity diagramming: group observations by similarity to surface patterns

Tips:

• Drop unimportant uncategorized data & keep only what serves your research goals.
• Beware of bias by analyzing your assumptions or having another researcher review your interpretations.
• Avoid flattening data into binary categories (positive/negative, yes/no).

Research report

A report includes 6 components:

• Research goals
• Product context: metrics or signals that flagged a problem and triggered the research
• Methodology & why use it
• Participant sample: size & profile of participants
• Insights & findings
• Recommendations

A goal tied to a real metric is much harder to ignore. E.g. a study about why users were abandoning their carts include the drop-off rate.

Key insights give busy stakeholders a short, scannable summary of the most important takeaways from your study. Detailed insights follow with the data and analysis that led to each takeaway. This is where you show the work: the patterns you identified, how you interpreted them, and what evidence supports your conclusions.

The recommendations section is where research earns its place in the product roadmap. Every recommendation must tie back to your data. If the data is not yet strong enough to support a definitive call, say so and recommend further research as the next step. That is still a useful output.

When your research spans multiple topics, structure the report by themes. For each theme, include a title, a short summary for readers who want the key takeaway fast, and a deeper section for those who want to understand the supporting data.

Personas

2 types of change that indicate the underlying data is going stale:

• Business and technology changes: users’ goals & tasks are going to change too.
• Changes in user behavior and demographics

→ Review and update personas

Most personas include 4 types of info:

• Name & image
• Demographics: age, location, occupation, education, technical skills
• Psychographics: goals, pain points, behaviors, motivations. Most important layer
• Tagline or quote: show attitude & primary goal

Most products need 3-7 personas. More means more maintenance & competing priorities. Distinguish between primary and secondary persona.

How to know if a detail is important to include in a persona? Ask if it would change a design decision.

Empathy maps

An empathy map shows what a user says, thinks, does, and feels. Can reflect a person or a pattern from multiple users.

Says examples:

• “I am loyal to ABC because I never have a bad experience.”
• “I want something reliable.”
• “I don’t understand what to do from here.”

Thinks examples:

• “This is really annoying.”
• “Am I dumb for not understanding this?”
• “This is taking too much time.”

Thinks cues: hesitations, facial expressions, body language, contradicting moments where users do different from what they say.

Does examples:

• Refresh a page several times
• Shop around to compare prices
• Check the size chart

Feels examples:

• Impatient because the pages load too slowly
• Confused by too many contradictory prices
• Worried about making a mistake

If you find the traditional 4-quadrant map too generic, add quadrants like Goals, Pain points, Tasks, Influences, Gains.

User journey & experience maps

A customer journey map captures a user’s process of accomplishing a goal with a product—what that person does, thinks and feels at each stage.

Most journey maps include: persona & scenario → phases (user actions, thoughts, emotions) → insights or opportunities. E.g. Jane Doe - looking for a new phone → Phases: consider - explore - compare - test - decide.

4 components of experience maps:

• Phases
• Actions: what people do to move forward
• Thoughts/mindsets
• Emotions

A journey map captures how users interact with a product. An experience map captures broader: what people think, feel, and do while solving problems.

Mental models

Mental models are the internal beliefs users bring to your product, not facts about how it actually works, but assumptions built from past experience.

Mental models are flexible. They shift as users encounter new products, pick up habits from other interfaces, and talk to people around them. That’s what makes them designable, not fixed traits you have to work around, but something you can actively support.

Because mental models are built on belief rather than fact, they are not always accurate. When you spot a mismatch, you can:

• Change the design to match how users think it should work.
• Help users form a more accurate model using clear labels, tooltips, contextual help, or onboarding.

The interface should be self-explanatory first and documentation should be a fallback.

Document user expectations in personas:

• Prior product experience: similar tools or interfaces they have used?
• Domain knowledge
• Tech literacy
• Cultural context

Notes from Uxcel UX Research course