This folder is a method selection handbook— covering 23 UX research methods, each explained thoroughly using the same nine-section template (overview / applicability / cases / steps / value / risks / comparisons / best practices / portfolio). But for PMs, the truly scarce resource is not 'how to execute each method', but"which method should I actually use for the problem in front of me". Therefore this article does not repeat the operational details of each method, but instead rebuilds the content around three things: ① It positions all 23 methods at once using the NNGroup four-quadrant framework; ② It provides three sets of selection tables organized by problem / stage / budget; ③ It groups the methods into five major families, explains 'when to use · how to choose · counterexamples' for each family, and maps out method combination chains(you rarely use a single method alone; real skill lies in combining methods).
It overlaps by approximately 35% with the existing NNGroup Qualitative Research and Research Methods Toolkit (method landscape / interviews / empathy maps / sample size / task design / moderation / expert evaluation / workshops / ethics) — interviews, focus groups, usability testing, and expert evaluation are covered in greater depth there, so this piece only provides pointers to those sections. The incremental value of this article lies in the other 13 methods (four quantitative methods / two experimental methods / two IA methods / Kano / service blueprint / journey map / field study / visual testing), as well as the 'method selection' perspective itself.
One-sentence core principle:Qualitative methods answer 'why / where the problem is', quantitative methods answer 'how big / how severe', expert methods answer 'let's do a cheap first pass to catch obvious issues', but no single method can deliver the full truth on its own — the essence of method selection is combination.
1. First, position them: laying out the 23 methods on the NNGroup four-quadrant framework
Any method can be positioned along two axes (NNGroup's classic framework):
| Axis | Two ends | Distinction |
|---|---|---|
| Nature of data | Attitudinal (what users say) ↔ Behavioral (what users do) | What people say and what they do are often opposite — don't rely only on asking if you want to know real behavior |
| Form of data | Qualitative (discover problems, provide direction) ↔ Quantitative (measure metrics, provide numbers) | Qualitative answers 'why / where', quantitative answers 'how severe / how many people' |
Placement of the 23 methods within the four quadrants:
| Quadrant | Methods | What it can tell you |
|---|---|---|
| Attitudinal + Qualitative | User interviews / focus groups / reaction card emotion words / visual design testing (preference questions) | Motivations, mental models, first impressions, brand perception |
| Behavioral + Qualitative | Usability testing / field studies / diary studies / task analysis / card sorting / tree testing / eyetracking | How users actually behave, where they get stuck, how they organize information |
| Attitudinal + Quantitative | Survey questionnaires (large sample) / Kano / visual testing (rating questions) | Satisfaction trends, distribution of need priorities, brand word hit rate |
| Behavioral + Quantitative | Analytics / conversion funnels / user paths (Sankey) / A/B testing / multivariate testing MVT | How many users drop off at which step, which version has higher conversion |
| Cross-quadrant frameworks | Root cause analysis RCA / journey maps / service blueprints | Not standalone methods, but containers/visual skeletons that 'hold other methods' |
Two sets of orthogonal tags:Formative(early stage, finding problems for improvement) vs Summative(evaluating overall performance, comparing against competitors/old versions);Remote vs In-person. Interviews/card sorting/tree testing/surveys can all be run remotely; field studies/eyetracking almost always require in-person presence.
2. Three sets of method selection decision tables (core of this article)
Select based on 'what question you want to answer'
| Your question | First-choice method | Why / counterexample |
|---|---|---|
| Who the users are, what their motivations/mental models are | User interviews (NNGroup Qualitative Research and Research Methods Toolkit (method landscape / interviews / empathy maps / sample size / task design / moderation / expert evaluation / workshops / ethics)) | Digs into 'why' during the exploration phase;Counterexample: don't use it to test whether an interface is usable |
| How users actually behave in real environments | Field Study / Diary Study | Reveals environmental interference and workarounds that you cannot see in a lab;Counterexample: high cost, not suitable for frequent validation |
| How many steps it takes to complete a goal, which steps have high cognitive load | Task Analysis (HTA hierarchical diagram) | Breaks down 'what they want' into 'how they do it step by step';Counterexample: don't only look at the current process, or you will carry inefficiencies into the new system |
| How content/navigation should be categorized (building from scratch) | Card Sorting | Reveals the categorization in users' mental models;Counterexample: it does not include page context, only provides top-level categories, and must be paired with tree testing for validation |
| How findable existing/candidate navigation is (validation) | Tree Testing | Tests findability using text-only menus, with success rate/direct hit rate metrics;Counterexample: it strips away visuals/search, so it does not equal real-world experience |
| How usable the interface is, where users get stuck | Usability testing (NNGroup Qualitative Research and Research Methods Toolkit (method landscape / interviews / empathy maps / sample size / task design / moderation / expert evaluation / workshops / ethics)) | The gold standard for behavioral+qualitative research; watch what users do before listening to what they say |
| Whether new users can learn it on first use | Cognitive Walkthrough (NNGroup Qualitative Research and Research Methods Toolkit (method landscape / interviews / empathy maps / sample size / task design / moderation / expert evaluation / workshops / ethics)) | Specifically evaluates learnability, no real users required |
| Tight budget, want to first scan for obvious pitfalls | Heuristic Evaluation (NNGroup Usability Testing and the Ten Usability Heuristics (full table of 10 heuristics · application to complex applications · user pleasure hierarchy · testing methods 5-user rule/recruitment/remote/task scenarios/step-by-step tasks/attitudinal vs behavioral/4-step analysis/small-sample error/competitive evaluation)) | 3–5 experts evaluate against the ten principles;Counterexample: it cannot replace real users |
| Large-sample attitudinal data/satisfaction/NPS | Survey (NNGroup Qualitative Research and Research Method Toolkit (Method Landscape / Interviews / Empathy Maps / Sample Size / Task Design / Moderation / Expert Evaluation / Workshops / Ethics)) | Fast, cheap, quantitative;Counterexample: only reflects attitudes, not behavior; many biases |
| Among a pile of features, which are must-haves / which are nice-to-haves | Kano Model | Classifies features into five categories using paired positive/negative questions;Counterexample: gives a satisfaction structure, not a to-do list |
| At which step users drop off, and how many drop off | Conversion Funnel | Staircase chart intuitively locates bottlenecks;Counterexample: only tells you "where people drop off" not "why" |
| The actual path users take (not the ideal path) | User Path Analysis Sankey | Captures detours / backtracking / multiple branches;Counterexample: the same path can correspond to completely different experiences; needs qualitative follow-up |
| Metrics suddenly show anomalies; systematically investigate root causes | Root Cause Analysis (RCA) | 5 Whys / fishbone diagram traces symptoms back to root causes;Counterexample: don't form a conclusion first then look for evidence (confirmation bias) |
| Which of two options to choose / whether a fine-tuning works | A/B Testing | Real traffic, reliable conclusions;Counterexample: only says "which is better" not "why"; hard to discover unexpected problems |
| The optimal combination of multiple elements on the same page | Multivariate Testing (MVT) | Tests all combinations at once and looks at interaction effects;Counterexample: consumes traffic; too many combinations make it impossible to run |
| Whether the visual style / brand feel is right | Visual Design Test / Reaction Card Emotion Words | 5-second test + controlled vocabulary quantifies "feelings";Counterexample: weak at judging impact on behavior |
| Whether users can read long-form content, what they ignore | Eyetracking | Physiological-level evidence of attention;Counterexample: "looked at" ≠ "understood"; equipment is expensive and sample requirements are high |
| Where cross-touchpoint / cross-department experience is fragmented | Journey Map → Service Blueprint | Start from the user perspective then move to the organizational perspective; trace UI problems back to organizational/process root causes |
| Continuous monitoring for high-traffic products; how big the problems are | Analytics for UX | Large-scale, real, continuous;Counterexample: looks at behavior not motivation; must be paired with qualitative research |
Choose by "which stage the project is in"
| Stage | Primary Methods | Guiding Principle |
|---|---|---|
| Exploration / Problem Definition | Interviews · Focus Groups · Field Studies · Diary Studies · Card Sorting · Task Analysis | This stage seeks divergence and understanding, almost entirely qualitative; don't rush into A/B testing |
| Concept / Information Architecture | Card Sorting → Tree Testing · Kano (Feature Prioritization) · Journey Map (Ideal Future State) | Polish the "skeleton" before visual drafts exist to reduce rework |
| Design Validation (Pre-Launch) | Usability Testing · Cognitive Walkthrough · Heuristic Evaluation · Tree Testing Retest · Visual Testing | Clear mines before launch: first use expert methods to sweep obvious pitfalls, then test deep pitfalls with real users |
| Post-Launch / Growth Optimization | Analytics · Funnel · Path · A/B · MVT · RCA · Diary Studies (Long-Term) | This stage seeks quantification and validation, primarily behavioral quantitative methods |
| Retrospective / Continuous Monitoring | RCA · Analytics Dashboards · Regular Lightweight Kano/Diary Studies | Pin the core funnel to dashboards, conduct routine retrospectives; don't only investigate when anomalies occur |
Choose by Budget / Capability Threshold (from cheap to expensive)
| Threshold | Methods | Notes |
|---|---|---|
| Nearly zero cost, can be done internally | Heuristic Evaluation · Cognitive Walkthrough · Task Analysis | Expert methods do not require recruiting real users |
| Low cost, requires recruiting a small number of users | Interviews (starting from 5–6 people) · Usability Testing (5 people) · Card Sorting (qualitative 15 people) · Tree Testing | For qualitative research, look for saturation; see NNGroup Qualitative Research and Research Method Toolkit (Method Landscape / Interviews / Empathy Maps / Sample Size / Task Design / Moderation / Expert Evaluation / Workshops / Ethics) |
| Medium cost, requires tools/platforms | Surveys · Focus Groups · Visual Testing · Remote Tree Testing/Card Sorting | Online tools (OptimalSort/Treejack/UserZoom) |
| High cost, requires traffic or equipment | A/B · MVT · Eyetracking · Field Studies · Long-Cycle Diary Studies | A/B requires sufficient traffic + time; eyetracking requires dedicated equipment; field studies require travel + long cycles |
| High labor, cross-department collaboration | Service Blueprint · Journey Map (Co-creation) · RCA | The cost is not money but "organizational time"; requires management support |
III. Detailed Explanation of the Five Families (When to Use / How to Choose / Counterexamples)
Family 1: Qualitative Exploration — Digging into "Why" and "Real Scenarios"
| Method | Core Positioning | When to choose it over neighboring methods |
|---|---|---|
| User Interviews | One-on-one deep dive into motivations/mental models | When you need individual in-depth stories; see details NNGroup Qualitative Research and Research Method Toolkit (Method Landscape / Interviews / Empathy Maps / Sample Size / Task Design / Moderation / Expert Evaluation / Workshops / Ethics) |
| Focus Group | Multi-person discussion to elicit opinions/attitudes | Use when you need mutual group stimulation and to see conflicts across different roles;Never use it to judge whether an interface is usable |
| Field Study | On-site observation in real environments | Environment/space/collaboration are key variables (hospitals, stores, factories, customer service centers) |
| Diary Study | Longitudinal, user self-recorded | Use when you need to observe "long-term, cross-day, gradual emotional changes" (learning curves, retention motivation) |
| Task Analysis | Break down "how it is done step by step" | For complex processes, to find high cognitive load/redundant steps, and produce an HTA hierarchy diagram |
Why this family: Behavioral data (Analytics/funnels) only tells you "40% of users dropped off at step 3", but never tells you "why they dropped off". Qualitative exploration fills this gap.
Classic counterexample for focus groups (must remember): NNGroup has repeatedly emphasized — when researchers demonstrated the "window model vs scrolling model" for text browsing, users in the group all verbally said they preferred scrolling, but in actual use the window model performed better. That is why modern GUIs use window logic. Lesson:Verbal preference ≠ actual performance; focus groups measure attitudes, not usability.
Hard numbers for diary studies: Cycle length is determined by behavior frequency (voice assistants used multiple times a day → 1 week; food delivery used several times a week → 2–3 weeks; choosing a phone completed within about 2 weeks). Longer is not better (longer cycles lead to high attrition). Three recording triggers: event-based (record every time the behavior occurs) / time-interval-based (fixed time every day) / signal-based (researcher pushes reminders). Recruit 5–12 / 12–30 people, over-recruit to guard against attrition; set a cap on rewards to prevent "fabricated diaries". After the study ends,be sure to conduct a follow-up interview, using the user's own diary as material for probing questions.
Unique value of field studies: See the full picture of "people-devices-environment" and "workarounds users cannot articulate". Common deliverables: journey map + service blueprint + field evidence card.
Family 2: Quantitative Behavior — Measuring "how big, how severe, which path they take"
| Method | What it looks at | Division of labor with neighboring methods |
|---|---|---|
| Analytics for UX | Macro: scale/trends/ongoing monitoring | Broadest scope, used for "problem indication"; see the five-dimensional positioning method below |
| Conversion Funnel | Conversion/drop-off at each step on a linear path | Answers "which step loses users", suitable for predefined "ideal paths" |
| User Path Sankey | Non-linear real paths, detours, backtracking | More flexible than funnels, can capture multiple branches; suitable for SaaS/content navigation |
| Eyetracking | Fixation points/scan paths (physiological-level attention) | The only method that can distinguish "whether they looked" rather than "whether they acted" |
Why this family: For high-traffic products, interviews cannot reach the full picture; behavioral data is a "telescope". But all four methods share the same caveat:Looking only at behavior cannot reveal motivation; it must be paired with qualitative research.
Five-dimensional positioning method for Analytics (practical): When metrics are anomalous, quickly triage causes across these five categories — ① Traffic (sudden drop in a certain channel) ② Technical (event not triggered/frontend error) ③ Content (unclear copy) ④ Navigation (no one clicks key links) ⑤ Visual (button is too weak). Pair with a measurement plan (goal → micro-conversion → metric → data source).
Funnel vs Path (do not confuse them): A funnel is "the staircase you predefined", a path is "the web users actually walk". Keep funnel levels at 4–7 steps; raw Sankey diagrams for paths have enormous noise,you must filter + highlight to make them readable. There is a counterintuitive trick for path analysis:Looking backward 3–5 steps from an endpoint (e.g., "delete account") reveals problems better than looking forward from the homepage.
Four eyetracking scan patterns (directly usable for content design): F-pattern (long paragraphs without hierarchy → large amounts of information ignored) / Spotted (skipping around to find keywords and lists) / Layer-cake (skimming along bold subheadings) / Commitment (high interest or studying for an exam → reading word by word). Design recommendations: add subheading hierarchy, use lists, bold keywords, keep paragraphs 1–3 lines, leave "landing points". Note that the F-pattern is mirrored for RTL languages such as Arabic.
Family 3: Controlled Experiments — "Which version is better"
| Method | What it tests | Selection rule |
|---|---|---|
| A/B Testing | Comparison of a small number of overall versions | Few changes, single goal, sufficient traffic |
| Multivariate Testing (MVT) | All combinations of multiple elements on the same page | When you want to know "which combination is optimal" + "the independent/interaction contribution of each element" |
Path dependence trap of A/B vs MVT (core knowledge point): Suppose you are optimizing two variables on a product detail page: "image/video" and "button copy".
- Sequential A/B: first test image vs video (video wins) → then test copy on the video version.Problem: If the true optimum is "image × Buy Now", this path will never detect it (the second round is locked into the "video" premise).
- MVT tests all 4 combinations at once, avoiding this local optimum, and can also separate main effects ("video overall is 15% higher than image") and interaction effects ("video × Add to Cart far outperforms others").
- Cost: The number of combinations multiplies as variables increase, consuming traffic. Start early practice with 2 variables × 2 variants.
Practical division of labor: Use A/B to set direction for major redesigns → use MVT for element-level refinement on the winning version. Classic A/B numbers: weakening the coupon input box → 20–50% sales lift; big button → +1% (but you won't know why).Both only tell you "which is better" not "why", and looking only at short-term results sacrifices long-term outcomes(stuffing search engines with ads increases short-term clicks but drives users away long-term). True 100%-level growth comes from major improvements in trust/information quality/core flows, which require qualitative discovery first.
Family 4: Expert Evaluation — No real users needed, sweep for issues cheaply first
| Method | Perspective | What it evaluates | Scope |
|---|---|---|---|
| Heuristic Evaluation | Analytical expert | Overall usability (against the ten principles) | Comprehensive |
| Cognitive Walkthrough | Simulating a novice | Learnability (ask 4 questions step by step) | Specific tasks |
Detailed operations (3–5 people evaluating independently, 5 severity levels, ten heuristics, 4 walkthrough questions) are thoroughly covered in NNGroup Usability Testing and the Ten Usability Heuristics (Full Table of 10 Heuristics · Complex Application Use · User Delight Hierarchy · 5-User Rule for Testing Methods/Recruitment/Remote/Task Scenarios/Step-by-Step Tasks/Attitude vs Behavior/4-Step Analysis/Small Sample Bias/Competitive Evaluation) and NNGroup Qualitative Research and Research Method Toolkit (Method Landscape / Interviews / Empathy Maps / Sample Size / Task Design / Moderation / Expert Evaluation / Workshops / Ethics); here we only provide selection key points:
- None can replace real users; they can only supplement and point directions for real-user testing.
- Low budget/early stage → run heuristic evaluation first to sweep obvious pitfalls, and save precious real-user time for deeper problems.
- Evaluating whether novices can get started on a key flow (registration/payment/account opening) → cognitive walkthrough.
- Heuristics are not laws: you may intentionally violate them (a hamburger menu on small screens violates "recognition over recall" but is reasonable due to space constraints), but mark it as "pending user research validation".
- For uncertain issues, it is better to mark "test this" than to make a call — an expert's personal preference is not objective evidence.
Family 5: Information Architecture (IA) — "How content is organized, whether users can find it"
| Method | What it does | Order |
|---|---|---|
| Card Sorting | Discover "what the structure should look like" (users group items bottom-up) | First |
| Tree Testing | Validate "whether this structure is findable" (find content in a text-only menu) | After |
Why use them as a pair: Card sorting reveals mental models (users group by semantic similarity, not by brand/format), but it does not include page context and only gives top-level categories; tree testing quantifies findability in a "pure menu" stripped of visuals/search.Standard pipeline: card sorting generates candidate structures → tree testing evaluates strengths and weaknesses → usability testing validates the overall real-interface experience.
Hard numbers (needed for selection):
- Card sorting sample size: qualitative ≥15 people, quantitative 30–50 people. Number of cards 30–50 (too many causes fatigue and creates a "miscellaneous" pile). NNGroup Fidelity study:15 people with r=0.90 is the optimal cost-effectiveness for most projects(5 people r=0.75 is insufficient, 30 people r=0.95, 60 people r=0.98 has extremely low cost-effectiveness). Three modes: open (free grouping, most commonly used) / closed (place into existing groups, for validation) / hybrid. Deliverables: similarity matrix / standardized grid / dendrogram.
- Core tree testing metrics:Task success rate + "direct hit rate" (correct on first choice vs multiple attempts). NNGroup example: for a task, 74% eventually found the right item, but only 50% found it directly (the other half had to go back and reselect) — so looking only at success rate is misleading. Task descriptions should be short and scenario-based,and must not directly use the original menu label words(otherwise success rate is artificially inflated). For multi-version comparison, use different participants for different trees (to avoid learning effects).
- Common IA pitfalls: users avoid vague broad categories such as "other/miscellaneous/general"; internal department jargon is completely unintelligible to external users; appropriately providing multiple entry paths reduces the cost of getting lost.
Family 6: Evaluation and Synthesis Frameworks — "Containers that hold other methods"
These are not single atomic methods, but frameworks/visual skeletons that hold other methods inside them:
| Framework | What it integrates | Key points |
|---|---|---|
| Usability Testing | Core evaluation combining behavior + qualitative | For task writing see NNGroup Qualitative Research and Research Method Toolkit (Method Landscape / Interviews / Empathy Maps / Sample Size / Task Design / Moderation / Expert Evaluation / Workshops / Ethics); in this article it is positioned downstream as "validating IA/funnel hypotheses" |
| Root Cause Analysis (RCA) | Funnel + path + session replay + interviews + A/B strung into a causal chain | Process: actionable problem statement → multi-source data → cohort/funnel/path localization → 5 Whys/fishbone diagram to generate hypotheses → validation → lock in root cause |
| Kano Model | Interviews (sorting features) + structured questionnaire (dual positive/negative questions) | Five categories: Must-be / One-dimensional / Attractive / Indifferent / Reverse. Attractive features sink into Must-be over time (touchscreens, free Wi-Fi) |
| Journey Map | Interviews/diaries/customer service/funnel data → a story for one persona in one scenario | 5 elements: persona + scenario expectations + stages (4–7) + behavior/thoughts/emotions + opportunities. The bottom must include "opportunities + owner + metrics" otherwise it cannot drive action |
| Service Blueprint | Expands the journey map downward (adding backstage/system/evidence swimlanes) | Use the "line of visibility" to separate frontstage/backstage; exposes issues that "look like UI problems but are actually organization/process problems" |
| Visual Design Testing | Combination of 5-second test + first click + preference + open/closed word list + rating + eyetracking + A/B | Measures "feel"; the word list must include positive + negative + distractor words, and negative words need to account for ≥40% to avoid distortion |
Actionable Kano formula: For each feature ask a positive question ("how would you feel if this feature existed") + a negative question ("how would you feel if it did not"), and classify using a matrix with five-level answers. You can optionally calculate two coefficients:Better = (Attractive + One-dimensional) / (Attractive + One-dimensional + Must-be + Indifferent) the higher it is, the more it increases satisfaction;Worse = −(Must-be + One-dimensional) / (same denominator) the closer to −1, the more it angers users when missing. Priority: Must-be + high Worse are mandatory (baseline) → high Better Attractive/One-dimensional for differentiation → cut Indifferent/Reverse. At least 30–50 people per segment.
RCA 5 Whys example: Starting from "new user Day 7 retention dropped", trace layer by layer to "onboarding flow too long / value point delayed / key feature hard to understand". The fishbone diagram groups hypotheses across five dimensions: "product experience / pricing strategy / market competitors / user types / technical stability". Cyberbiz practice: the root cause of low feature adoption was not technical but usability and learning cost (entry position/unclear naming, too many jumps in the path).
Reaction Card emotion word method (Microsoft Reaction Card): Screen 20–25 visual-related words from Microsoft’s 118-word list, and ask users to select the 5 that best describe the interface. Remove function-related words (slow/technical) and keep visual-related words (fresh/calm/professional). Randomize the order in online surveys to prevent order bias. Output a word-frequency bar chart plus a two-group Venn diagram (differences between younger vs. older users).
IV. Method Combination Chains (Rarely Used Alone)
The highest level of method selection is not picking a single method, but designing a chain of "discover → validate → iterate." Several classic chains repeatedly demonstrated by NNGroup:
| Scenario | Chain |
|---|---|
| Information architecture redesign | Card sorting (discover structure) → Tree testing (validate findability) → Usability testing (overall experience of the real interface) |
| Long-term experience understanding | Interviews (current mental models) → Diary study (2-week longitudinal) → Follow-up interviews (use diaries to probe "why they persisted/stopped") |
| Conversion rate diagnosis | Funnel (identify which step drops users) → Path analysis (examine detours) → Session replay + heatmaps (observe specific interactions) → Interviews/surveys (ask about feelings) → A/B testing (validate changes) |
| Metric anomaly troubleshooting | Coordinated under the RCA framework: Analytics signals → Cohort/funnel/path localization → 5 Whys to generate hypotheses → Usability testing/session replay/small A/B tests for validation → Lock in root cause and produce solutions |
| Service experience upgrade | Field studies/interviews (discover pain points) → Journey map (story from the user’s perspective) → Service blueprint (add backstage processes) → Future blueprint comparison → Action checklist |
| Feature prioritization | Interviews (sort out candidate features) → Kano survey (classify into five categories + Better/Worse scores) → Combine with cost/technology/regulatory factors → Roadmap |
| Visual direction decision | 5-second test (first impression) → Reaction cards/closed word list (brand word hit rate) → Preference test (choose one among 2–3 versions) → A/B testing (validate clicks after launch) |
General principles (applicable across all methods):
- Attitudinal data must be calibrated with behavioral data——Users say "I don’t care about shipping fees," yet repeatedly compare shipping fees at checkout (empirically demonstrated by combining interviews + usability testing).
- Quantitative methods find "where," qualitative methods explain "why"——When a funnel shows drop-offs, always pair it with at least one qualitative method (session replay/interviews/on-site observation).
- Frequent small studies > one large, complex study(Diminishing marginal returns).
- All behavioral data depends on the quality of event tracking——Inconsistent naming/missed tracking points → "professional-looking but wrong conclusions."
Sourced from NNGroup Topics / UX Research Methods (23 articles). In-depth coverage of interviews/focus groups/usability testing/expert evaluation/workshops/ethics/sample size can be found in NNGroup Qualitative Research and Research Methods Toolkit (Method Landscape / Interviews / Empathy Maps / Sample Size / Task Design / Moderation / Expert Evaluation / Workshops / Ethics)(approximately 35% overlap; this article focuses on method selection and the other 13 methods); for quantitative methods, see NNGroup Quantitative Research (A/B Testing · Sample Size · Confidence Intervals · UX Metrics · Conversion Analysis · Benchmarking · Learning Curves); for the ten heuristics and usability, see NNGroup Usability Testing and the Ten Usability Heuristics (Complete Table of 10 Heuristics · Application to Complex Applications · User Pleasure Hierarchy · Testing Methods: 5-User Rule/Recruiting/Remote/Task Scenarios/Step-by-Step Tasks/Attitude vs. Behavior/4-Step Analysis/Small-Sample Errors/Competitive Evaluation); complementary to Google UX Certificate / UX Research and Empathize-Define. Images/PDFs (such as the heuristics workbook) are retained in sources.
Sources and Related Materials
- NNGroup Qualitative Research and Research Methods Toolkit (Method Landscape / Interviews / Empathy Maps / Sample Size / Task Design / Moderation / Expert Evaluation / Workshops / Ethics)
- NNGroup Quantitative Research (A/B Testing · Sample Size · Confidence Intervals · UX Metrics · Conversion Analysis · Benchmarking · Learning Curves)
- NNGroup Usability Testing and the Ten Usability Heuristics (Complete Table of 10 Heuristics · Application to Complex Applications · User Pleasure Hierarchy · Testing Methods: 5-User Rule/Recruiting/Remote/Task Scenarios/Step-by-Step Tasks/Attitude vs. Behavior/4-Step Analysis/Small-Sample Errors/Competitive Evaluation)
- Google UX Certificate / UX Research and Empathize-Define