Combine Course 2 (Start the UX Design Process: Empathize → Define → Ideate) and Course 4 (Conduct UX Research and Test Early Concepts) into one "research craft handbook".Basic Concepts Those courses already covered the 5 stages of design thinking and the three types of research; here we only cover operations— how to conduct interviews, how to draw empathy maps, how to write problem statements, how to run usability tests, how to turn a pile of sticky notes into insights, and how to present insights to your boss.
Significance for PMs: This course and the Yitang user interviews you are familiar with (local reference material), OSCAR research (local reference material)are two ways of describing the same thing. The difference: Yitang leans toward "business decision research" (whether to pursue this business), while UX leans toward "product interaction research" (where to place this button).Usability studies and affinity diagrams are two UX-exclusive tools that PMs would do well to borrow..
I. UX Research Landscape: Three Types × Timing
Research objective = Understand user behaviors, needs, and motivations through observation + feedback, so that design is based on facts rather than assumptions — bridging the gap between "what the company thinks users want" and "what users actually want". Research runs through the entire process, not just at the beginning. The three types (see details in Google UX Certificate / UX Design Fundamentals §9):
| Type | Timing | Core Question |
|---|---|---|
| Foundational Research | Before design | What should we build?? |
| Design Research | During design | How should we build it? |
| Post-Launch Research | After release | Were we successful? |
3 traits of a good researcher:Empathy(standing in the user's shoes),Pragmatism(focusing on feasibility and outcomes, helping the team stay aligned on goals),Collaboration.
Instructor Case · The Power of Research (Emily, Senior Researcher, Google Assistant): She repeatedly told engineers "Children may not be able to read yet; this feature has to be changed," but the engineers just didn't understand (two different knowledge systems). She simply invited the engineers to come on site and watch a 7-year-old girl use the App — the girl got stuck the whole way, and finally blurted out "I can't read". All the engineers instantly gained a whole new level of empathy for "research" and "users".Bringing the team to the research site in person > repeatedly explaining the logic(the persuasive power of qualitative research).
II. Research Plan: 7 Elements
Before formal research, write a Research Plan to align the team's understanding. 7 elements:
- Project Background: Why you are doing this research.
- Research Goals: Simple, clear, and directly pointing to the problem. Ask yourself: Who are the users? What do you want to understand? Which pain points need to be dug into?
- Research Questions: The specific questions you need to answer.
- Key Performance Indicators (KPIs): Quantitative metrics to measure whether goals are met (e.g., task completion time, number of clicks, success rate, adoption rate, retention).
- Methodology: What methods to use and how to execute them.
- Participants: Target user characteristics (age/location/occupation/gender/whether they have disabilities); recruit diversely to ensure representativeness.
- Script: A list of questions read verbatim during interviews/tests to ensure everyone is asked the same way.
Privacy and Ethics(runs throughout): Sign non-disclosure agreements (NDAs) to protect unreleased designs + users' personal information; treat privacy as part of UX practice, not an afterthought.
Instructor Perspective · Ethics Runs Through Every Step (Vanessa, Google Researcher): Ethics/privacy is not just one single step — from selecting who to interview(include historically marginalized voices; samples should represent global users), to recruiting/interviewing/giving recommendations/interpreting/sharing/building the product,every step can "create good outcomes or cause accidental harm". The yardstick for handling user data:"If this were my data, would I want to be treated this way?"
III. Empathize Stage Craft
User Interview SOP (5 Steps)
graph LR
P["1 准备<br/>脚本/工具/调研用户/练习"] --> M["2 接待<br/>破冰/感谢/收基础信息"] --> C["3 进行<br/>开放式+追问+礼貌倾听"] --> N["4 记录<br/>金句/肢体语言/录音"] --> E["5 结束<br/>问补充/再感谢/留好印象"]
- Prepare: Write the script (ask everyone the same way), prepare tools (audio recording/video recording/video conferencing), check basic user information (to keep the conversation natural; in corporate settings also confirm they are not from competitors), and practice timing in advance.
- Conduct: Ask open-ended questions(not "Do you like going to the library?", but "Why do you like/dislike going to the library?") + follow-up questions("Can you tell me more?" "Why did that happen?").
- Record: Note verbatim golden quotes(that reflect real thoughts), body language, and (with consent) record audio and video.
- Keep the user comfortable throughout — interviews are for building empathy + capturing real needs, not interrogation.
Three Preparations + 5 Traits of Interview Questions
Define three things before research:research objectives(what problem to solve),target user characteristics, interview questions. 5 traits of good interview questions: ① Relevant (aligned with research objectives) ② Open-ended ③ Simple and clear ④ Neutral(not leading) ⑤ Allows follow-up.
Counterexample: "Do you order takeout because you don't have time to cook, or because you're too lazy to cook?" — this already assumes the reason for the user and limits their answers. Change it to "What difficulties have you encountered when ordering takeout?"
Empathy exercise before interviews: Before formal interviews, first simulate how a user (e.g., the fictional persona Reena, a paramedic with dyslexia) would answer, and discover in advance that words like "difficulty" might make people uncomfortable, then change to a friendlier phrasing.
Empathy Map
Break interviews into 4 quadrants —Says / Thinks / Does / Feels. Two types: single-user map / aggregated map (combine multiple people to find commonalities).
| Quadrant | What to fill in | Notes |
|---|---|---|
| Says | User's verbatim words | Do not summarize or rewrite on your own |
| Thinks | User's thoughts | Do not over-guess |
| Does | Actual behaviors | Record objectively |
| Feels | Emotions | If unsure, ask directly "How did this make you feel?" |
Value: Discover problems / inspire design / identify user groups (laying the foundation for building personas and segmenting markets).
Persona (≠ User Profile)
- User Profile: A short fictional bio, used for informal empathy exercises.
- Persona: A typical user group(not a single real person) summarized after a large number of interviews, guiding the entire design process. It turns complex information into memorable characters, keeps the team thinking from the user's perspective at all times, and makes it easier to explain decisions to stakeholders.
IV. Define Stage Craft: From Pain Points to Value Proposition
Logic chain:User stories/journey maps → pain points and emotions → problem statement → hypothesis statement → value proposition.
Problem Statement
A short piece of text describing user needs. 3 criteria for a good problem statement:Human-centered / Appropriately scoped(broad enough to leave room for creativity, specific enough to be actionable) / Actionable.
Formula:
[用户名] 是 [用户特征],需要 [需求],因为 [痛点/关键发现]。Example: Arnold is a busy corporate executive who needs a simple, easy-to-use app to hire dog walkers because he is not very familiar with technology.
Use 5W1H to clarify the problem: Who / What (the problem) / Where (used where) / When (when it happens) / Why (why it matters) / How (how they want to achieve it). The 4 values of a problem statement: clarify goals, expose constraints, determine the final product form, and set measurable success criteria.
Hypothesis Statement
The bridge after defining the problem and before ideating solutions. It is a reasonable guess based on research — "We believe a certain design can solve a certain need." Two writing formats:
| Structure | Perspective | Example |
|---|---|---|
| If / Then | User perspective, easy to test | If Arnold downloads the dog-walking app, then he can use "Easy Mode" to view only basic features |
| We Believe | Team perspective, builds consensus | We believe that providing Easy Mode for Arnold will help him hire dog walkers more easily |
One problem can have multiple hypotheses; hypotheses are temporary, and are validated, adjusted, or overturned as testing proceeds — this is the core of design thinking.
Value Proposition
The main value + unique benefits the product provides to users, answering "what the product does" + "why users should care". It is a synthesis of multiple hypothesis statements. 4-step method:
- List all features and benefits(quantity first, don't filter yet).
- Filter out the truly valuable ones(use interviews to find the categories users care about most, such as ease of use/professionalism/cost/reliability, and delete irrelevant features).
- Pair features ↔ user needs(every value proposition corresponds to a real need).
- Confirm the final list, then compare against competitors' negative reviews to find unique value propositions (competitors are often complained about for "insufficient dog walker training" → we highlight "professionally trained dog walkers with strict standards").
Case: When Gmail launched in 2004, it broke through with two unique value propositions — conversation view for emails + 1GB of free storage (1000 times more than competitors).
V. Human Factors Engineering and Design Psychology
Human Factors = The psychological/behavioral characteristics users bring to the product (impatience, limited memory, limited attention, error-proneness, bias, need for motivation…). History: WWI aviation psychology sought to "select people to adapt to machines"; WWII, because personnel backgrounds were too diverse, shifted to "making machines adapt to all kinds of people" — modern ergonomics was born.TL;DR (Too Long; Didn't Read) is a classic design that caters to "impatience + limited attention + limited memory".
3 core psychology laws(PMs can apply directly):
| Law | Meaning | Design Application |
|---|---|---|
| Von Restorff / Isolation Effect | Among a group of similar items, the one that is different is remembered most | Use contrasting colors to highlight CTA buttons (the bright blue "Start" button on maps) |
| Serial Position Effect | People remember the beginning (primacy) and end (recency) best; the middle is easy to forget | Place important functions at both ends of the navigation bar / at the beginning and end of long lists |
| Hick's Law | The more options there are, the slower the decision | Reduce options on a single screen, guide step by step, provide "recommended/default" options |
Two other basic concepts:Mental Model(users' expectations of "how things should work"; design should match expectations, e.g., a door should look like it can be opened),Feedback Loop(give result information after an operation; positive feedback builds confidence, negative feedback without explanation makes people give up).Color psychology: Red = dominance/warning/delete, Blue = safe and trustworthy, Green = confirm success.
Ethical bottom line: Use psychology to help users achieve their goals, not to manipulate. Do not exploit psychological weaknesses (dark patterns), do not create anxiety/addiction.
VI. Usability Study (UX's Signature Tool)
Let users actually operate the prototype/product, and observe where they get stuck — can be done at the early sketch stage, prototype stage (most commonly used), or after launch. Usually 5 users are enough to find the main problems and save cost.
Moderated vs Unmoderated
| Moderated | Unmoderated | |
|---|---|---|
| Format | Researcher is present (in-person/video) to guide | Users complete tasks independently; the whole process is recorded and reviewed afterward |
| Pros | Builds trust, real-time follow-up questions, guides key tasks, suitable for complex flows/sensitive topics, controllable conversation | More authentic and natural, larger sample size, flexible scheduling, lower cost, stronger sense of privacy, suitable for simple flows |
| Cons | Possible leading bias, hard to schedule, high cost, dependent on network, limited population coverage | No real-time support, cannot ask follow-ups, uncontrollable environment, intellectual property risk, platform may not be accessible, messy data hard to analyze |
Which one to choose depends on the research goal, user characteristics, process complexity, and available resources.
Best practices: Before / During / After
Before testing: ① Build rapport with the participant (small talk to help them relax / express sincere gratitude / explain the value of their feedback) ② Reassure them to speak honestly ("If you can't complete a task, it's the product's fault, not yours", and negative feedback will not offend the team) ③ Sign an NDA ④ Explain the process clearly ("Please think aloud as you complete tasks" / "This is not a test; there are no right or wrong answers") ⑤ Adjust your own mindset to stay neutral.
During testing 3 communication strategies: ① Ask more open-ended questions (e.g., "What do you think of this step?" instead of "Do you like this?") ② If you don't understand something, keep asking follow-up questions instead of making assumptions ③ Use "paraphrase to confirm" only at the end (e.g., "You mean filling in the address is too cumbersome, is that right?") — using it too early will lead the user.
Take notes (even if there is a recording): Note real-time observations (facial expressions / tone / word choice), summarize overall impressions, and extract verbatim golden quotes. You can use pen and paper / sticky notes (one observation per note, convenient for later affinity diagramming) / spreadsheets (for multi-person collaboration). At Google,multiple people observe and take notes at the same time, then consolidate the notes after the session.
Biases in usability research
| Source | Bias | How to address it |
|---|---|---|
| Moderator | Implicit bias / serial position effect | Stay neutral, cover the full process in notes |
| Participant | Friendliness bias(saying nice things to please you) | Maintain a professional relationship, repeatedly emphasize the need for honest feedback |
| Participant | Social desirability bias(saying favorable things to seek approval) | Provide multiple viewpoint options for them to choose from, emphasize "there are no right or wrong answers" |
For the full 6 types of research biases (confirmation / false consensus / recency / primacy / implicit / sunk cost) + 8 anti-bias practices for data collection, see Google UX Certificate / UX Design Fundamentals §9.
VII. Analysis and synthesis: From data to insights
Insight = An important finding about users that helps you understand users from a new perspective and guides next steps. Generated in 4 steps:
graph LR
G["1 汇总数据<br/>所有便签/表格/录音集中"] --> O["2 整理<br/>亲和图分组"] --> T["3 找共同点<br/>跨用户的模式"] --> I["4 每个主题写洞察<br/>问题+行为+设计调整"]
Affinity Diagram
"Affinity" = having common ground. Group similar observations together to identify common problems. Method: ① Write each observation on one sticky note(one clear sentence only, put verbatim user quotes in quotation marks) ② Stick all notes on a whiteboard/wall ③ Compare notes one by one and group similar ones together (like organizing a bookshelf) ④ Step back to look at the whole picture, adjust groups, and leave isolated notes separate for now ⑤ End up with 3–10 groups; if there are too many, split into subgroups. Digital tools: Jamboard / Miro / Mural / FigJam.
Accessibility tip: Visually impaired team members can also participate — have observations read aloud, others write them down and read them back, and let the visually impaired member decide how to group them.
Patterns → Themes
Don't stop at grouping; think about what users are actually experiencing. From the grouped sets, write down feelings and needs in plain language from the user's perspective. Example: Pattern "4 users said they couldn't find the confirmation page" → Theme "Users want to double-check information before placing an order to avoid mistakes". Tip: Mark uncertain items as "pending"; use colors/numbers to label different participants to see which points are raised by the majority.
5 criteria for strong insights
| Criterion | Explanation |
|---|---|
| Supported by real data | Comes from actual feedback from multiple participants (e.g., 3 out of 4 people), not subjective feelings |
| Answers the research question | Corresponds to the questions in the research plan, so stakeholders understand why it matters |
| Concise and clear language | People who did not participate in the research can understand it, no jargon overload |
| Evokes empathy | Helps the team understand users' emotions and difficulties, and motivates them to make improvements |
| Points to specific actions | Doesn't just state problems, but also suggests how to optimize (✅ "Users want to see all details before placing an order" ❌ "This app is easy to use") |
VIII. Sharing insights: 7 presentation tips
After completing research, you need to present it to the team/stakeholders. 7 tips:
- Know your audience: For marketing teams, talk about user behavior + business value; for management, talk about how to achieve company goals.
- Start with an overview: Put a table of contents at the beginning + summary page (state conclusions first) — present findings and impact first, then go into details.
- Tell stories: Have a beginning and end, add verbatim user quotes / photos / videos, so people see the product from real users' perspectives.
- Show data: Use charts instead of plain text to show trends faster and be more persuasive.
- Less text is better: Slides should only show the most important points (why / what / how); the audience should listen to you, not read the PPT. Put details in the appendix.
- Give clear next-step recommendations: How to modify the product, what additional research to do, what meetings to hold.
- Leave time for questions: Let the audience clarify points, dive deeper, and give feedback.
Supporting soft skills: Overcome stage fright, learn from excellent public speakers. Design critique is a high-frequency scenario for this set of presentation skills — see details in Google UX Certificate / Testing · Iteration · Design Critique · Delivery.
Glossary of key terms
| English | Chinese | One-sentence explanation |
|---|---|---|
| Empathy Map | Empathy map | 4 quadrants: Says / Thinks / Does / Feels |
| User Profile / Persona | User profile / Persona | Mini biography (informal) / Typical user group (guides the whole process) |
| Problem Statement | Problem statement | [User] needs [need] because [pain point] |
| Hypothesis Statement | Hypothesis statement | If-Then / We Believe, a testable guess about a solution |
| Value Proposition | Value proposition | The unique value of a product, refined via a 4-step method |
| 5W1H | — | Who/What/Where/When/Why/How to clarify the problem |
| Human Factors | Human factors | Psychological and behavioral characteristics users bring to the product |
| Von Restorff / Hick's Law / Serial Position | Isolation effect / Hick's law / Serial position effect | Three applicable psychology laws |
| Mental Model | Mental model | Users' expectations of "how something should work" |
| Moderated / Unmoderated | Moderated / Unmoderated | Two forms of usability studies |
| Think-aloud | Think-aloud | Having users verbalize their thought process while operating |
| Affinity Diagram | Affinity diagram | Grouping sticky notes by similarity to find commonalities |
| Insight | Insight | An action-oriented finding about users |
| KPI / NDA | Key Performance Indicator / Non-Disclosure Agreement | Quantitative metrics / Protect unreleased designs |
<!-- Video gap filled: Vanessa (ethics/privacy) + Emily (power of research) transcribed via Feishu Miaoji on 2026-06-02, key points incorporated into §1/§2. Verbatim transcripts stored in sources/.../_minutes/. The 5 Mock Usability Study sessions (Participant A-E) are for demonstration and not transcribed; methods are covered above. -->