{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-05-31T19:07:27.341Z"},"content":[{"type":"documentation","id":"e941cc19-dda0-4fd1-9b2b-f332a1200746","slug":"design-thinking-research","title":"Design Thinking Research: The Complete Guide to the Empathize Phase","url":"https://www.koji.so/docs/design-thinking-research","summary":"Design thinking is a 5-stage human-centered framework (Empathize, Define, Ideate, Prototype, Test) where research-driven empathy underpins all design decisions. Companies using design thinking outperformed the S&P 500 by 211% over 10 years. The Empathize phase uses user interviews, contextual inquiry, observation, and empathy maps to discover real user needs. Insights are synthesized into \"How Might We\" statements that drive ideation. AI-moderated interview platforms can reduce empathy research time by up to 44% while scaling to hundreds of participants.","content":"## Why the Best Products Start With Empathy — Not Ideas\n\nThe most expensive mistake in product development isn't bad execution. It's building the right solution to the wrong problem.\n\nDesign thinking exists to prevent exactly that. It's a human-centered framework that starts not with ideas, not with technology, but with deep understanding of the people you're designing for. And its first and most critical phase — Empathize — is fundamentally a research discipline.\n\nThe numbers behind design thinking are striking:\n\n- **Design-led companies outperformed the S&P 500 by 211% over 10 years** (Design Management Institute)\n- Organizations with mature design thinking practices deliver a **median per-project ROI of 229%** (Forrester Research)\n- IBM clients implementing design thinking saw **$48.4M in benefits over three years against $12M in costs** — a 301% ROI (Forrester)\n- **71% of companies** report that design thinking improved their working culture; **69%** say it made innovation processes more efficient (IxDF)\n\nNone of these outcomes happen without rigorous empathy research at the front end.\n\n---\n\n## What Is Design Thinking?\n\nDesign thinking is a problem-solving methodology that prioritizes deep human understanding before any solution is considered. Stanford's d.school, IDEO, and IBM have each developed influential versions, but the core structure is consistent across interpretations.\n\n> \"Design thinking is a human-centered approach to innovation that draws from the designer's toolkit to integrate the needs of people, the possibilities of technology, and the requirements for business success.\"\n> — Tim Brown, CEO of IDEO\n\nThe classic framework has five stages:\n\n1. **Empathize** — Understand the people you're designing for through research\n2. **Define** — Synthesize research into a clear problem statement\n3. **Ideate** — Generate a wide range of possible solutions\n4. **Prototype** — Build quick, low-cost representations of ideas\n5. **Test** — Gather feedback on prototypes with real users\n\nA critical misconception: these are not sequential steps. Design thinking is explicitly iterative. Teams regularly loop back — new findings in the Test phase often return them to Empathize, or new prototypes reveal the need to Redefine the problem. The diagram of a linear process understates the framework's organic, recursive nature.\n\n---\n\n## The Empathize Phase: Why It's the Foundation\n\nThe Empathize phase is where all subsequent design decisions are grounded. Skip it or shortchange it, and every downstream step is built on assumption rather than evidence.\n\n> \"Build bridges of insight through empathy — see the world through the eyes of others, understand the world through their experiences, and feel the world through their emotions.\"\n> — Tim Brown, IDEO\n\nEmpathy in design thinking is not sympathy (feeling *for* someone) — it's a rigorous effort to understand a person's experience, motivations, and the meaning they make of their world. The Empathize phase is what separates design thinking from traditional product development, which often begins with a solution looking for validation.\n\nThe phase has four research stages:\n\n1. **Discovery** — Broad exploration to understand the problem space and user context\n2. **Immersion** — Deeper engagement to experience the problem firsthand\n3. **Connection** — Identifying patterns and emotional resonance across research data\n4. **Detachment** — Stepping back to synthesize insights objectively\n\n---\n\n## Core Empathy Research Methods\n\n### 1. In-Depth User Interviews\n\nUser interviews are the cornerstone of the Empathize phase. Unlike surveys or analytics, they surface the \"why\" behind behavior — motivations, frustrations, mental models, and unarticulated needs that no other method can reach.\n\n**Effective empathy interviews:**\n- Ask about behavior and experience, not opinions: \"Tell me about the last time you tried to do this...\" not \"Do you think this is important?\"\n- Use open-ended questions that invite narrative: \"Walk me through a typical day when this comes up\"\n- Probe for emotional context: \"How did that make you feel?\" \"What was frustrating about that?\"\n- Listen more than you talk — the 80/20 rule applies\n- Follow surprising threads rather than sticking rigidly to a script\n\nFor the Empathize phase specifically, exploratory and generative interview approaches work better than structured validation interviews. You're not testing a hypothesis — you're understanding a human experience.\n\n### 2. Observation and Contextual Inquiry\n\nPeople don't always do what they say, and they don't always know what they actually do. Direct observation in context captures behavior that interviews cannot.\n\n**Observation techniques:**\n- **Naturalistic observation:** Watch users in their natural environment without interference\n- **Contextual inquiry:** Combine observation with concurrent interviewing — \"I notice you just paused there. What were you thinking?\"\n- **Shadowing:** Follow a user through their entire workflow for hours or days\n- **Fly-on-the-wall:** Observe without interacting; capture behavioral patterns over time\n\nContextual inquiry is particularly valuable for identifying the gap between stated preference and actual behavior — one of the most productive tensions in user research.\n\n### 3. Immersion Studies\n\nImmersion means putting yourself in the user's position as directly as possible. A healthcare team designing patient room layouts might spend 24 hours as a patient. A team redesigning a banking interface might attempt to open an account themselves using only the current system.\n\nImmersion produces visceral empathy that interviews alone cannot generate. It also surfaces operational details and edge cases that users have normalized to the point of forgetting to mention.\n\n### 4. Surveys and Diary Studies\n\nFor broader reach or longitudinal insights, surveys and diary studies complement the depth of individual interviews.\n\n**Diary studies** ask participants to log their experiences over days or weeks — capturing moments as they happen rather than relying on recall. They're especially valuable when the experience you're researching is intermittent or happens in contexts you can't observe directly.\n\n**Empathy surveys** are best used to quantify patterns discovered in qualitative work — how common is this frustration? How many people experience this workflow? Koji's [structured questions](/docs/structured-questions-guide) enable this quantification within conversational AI interviews, combining the depth of open-ended dialogue with the measurability of scale, ranking, and choice questions.\n\n### 5. Listening Sessions and Workshops\n\nGroup listening sessions (not focus groups, which introduce social dynamics that distort individual perspectives) bring 3–5 users together to surface collective patterns in experience. Facilitated workshops where participants map their own journeys, draw their mental models, or role-play scenarios often yield insights that one-on-one interviews miss.\n\n---\n\n## Empathy Mapping: Synthesizing What You Learn\n\nThe empathy map is the primary synthesis tool of the Empathize phase. It organizes research findings around four quadrants:\n\n| Quadrant | What You're Capturing |\n|---|---|\n| **Says** | Direct quotes from interviews and observations |\n| **Thinks** | What they believe — often not said aloud |\n| **Does** | Observable behaviors and actions |\n| **Feels** | Emotional states, worries, aspirations |\n\nBelow the quadrants, add:\n- **Pains:** Frustrations, obstacles, and fears\n- **Gains:** Goals, desires, and measures of success\n\n**How to build an empathy map:**\n1. Gather your research team and all research artifacts (transcripts, observation notes, photos)\n2. On sticky notes, capture every observation that fits a quadrant\n3. Cluster similar observations within each quadrant\n4. Identify tensions and contradictions (e.g., Says one thing, Does another)\n5. Synthesize into 3–5 core insight statements that capture the user's experience\n\nThe tensions are often the most generative insights. A user who *says* they want more control over the interface but *does* use only default settings is telling you something important about the gap between desire and behavior.\n\n---\n\n## From Empathy to Define: \"How Might We\" Statements\n\nThe Define phase begins where the Empathize phase ends. The primary tool for transitioning between them is the **\"How Might We\" (HMW) statement** — a reframing technique that transforms research observations into actionable design opportunities.\n\nHMW was originated by Procter & Gamble in the 1970s and later popularized by IDEO and the Stanford d.school. The format is deliberately constrained:\n\n**Structure:** \"How might we [verb] [object] so that [outcome]?\"\n\nThe three-word opener does specific work:\n- **\"How\"** acknowledges that a solution exists, projecting optimism\n- **\"Might\"** creates permission to explore — it's not \"how will we\" (too committing) or \"how should we\" (too prescriptive)\n- **\"We\"** establishes collective ownership of the problem\n\n**Examples of HMW statements derived from empathy research:**\n\n*Research finding: Users feel overwhelmed when they first open the dashboard and don't know where to start.*\n→ HMW make first-time users feel confident immediately upon login?\n\n*Research finding: Users abandon checkout when they realize they can't go back to change an earlier step.*\n→ HMW give users a sense of control and safety throughout the purchase flow?\n\n*Research finding: Users feel embarrassed asking for help because it signals to colleagues that they don't understand the product.*\n→ HMW enable users to get help discreetly without social exposure?\n\nWell-formed HMW statements are broad enough to invite multiple solutions but specific enough to exclude irrelevant ones. They embed the empathy insight directly into the design challenge.\n\n---\n\n## Common Empathy Research Mistakes\n\n### \"We already know our users\"\nThe most dangerous assumption in product development. Familiarity with your users is not the same as empathy. Teams who skip the Empathize phase consistently build solutions to the problems they imagine users have — not the problems users actually experience. Industry data suggests most companies that skip systematic empathy research are building on assumptions that are 40–60% wrong.\n\n### Confusing empathy with sympathy\nSympathy is feeling *for* someone. Empathy is understanding *from* their perspective. A researcher who feels sorry for a user struggling with an interface is being sympathetic. A researcher who understands *why* the interface feels confusing from that user's mental model is practicing empathy. The distinction matters for what you do with the insight.\n\n### Asking about opinions instead of behavior\nThe most common interview mistake in the Empathize phase: asking \"What do you think of [feature]?\" instead of \"Tell me about the last time you tried to [accomplish goal].\" Opinions are unstable and context-dependent. Behavioral narratives reveal durable patterns.\n\n### Following a rigid script\nEmpathy research rewards curiosity. When a participant says something unexpected, that's signal — follow it. A strict adherence to a pre-written question list produces surface-level data and misses the surprising insights that drive breakthrough solutions.\n\n### Insufficient sample diversity\nResearching only your most vocal or most satisfied users creates a false picture of the experience. The most actionable insights often come from edge cases, frustrated users, or people who have churned — people who feel the problem most acutely.\n\n### Skipping secondary research\nPrimary empathy research should be informed by existing data: analytics, support tickets, NPS verbatims, social media. Secondary research identifies where the friction is before you talk to users, making primary research more targeted and productive.\n\n---\n\n## How AI Is Transforming the Empathize Phase\n\nTraditional empathy research has a fundamental scaling problem. In-depth interviews are expensive, scheduling is painful, and geographic constraints limit who you can talk to. A typical empathy research sprint — 12–15 interviews — takes 4–6 weeks and costs $15,000–$30,000 in researcher time.\n\nAI-moderated interview platforms are changing this economics fundamentally:\n\n**Scale:** Where a team could run 15 interviews in a month, AI platforms can conduct 200–500 interviews in 72 hours — across time zones, languages, and user segments simultaneously.\n\n**Consistency:** AI interviewers never get tired, never ask leading questions, and never skip follow-up probes. They apply the same empathy interview protocol to every participant with perfect consistency.\n\n**Depth:** Voice-based AI interviews produce conversational richness comparable to human moderation. When participants hint at something emotionally significant, the AI probes deeper — the same way a skilled human interviewer would, but at scale.\n\n**Synthesis:** Automatic thematic analysis across all conversations surfaces the patterns that define the Define phase without weeks of manual coding. Platforms like Koji identify emerging themes across hundreds of transcripts and surface the most cited insights with direct quotes.\n\n> \"Leveraging AI in developing empathy maps can reduce research time by up to 44%, while machine learning algorithms identify patterns that human researchers might miss.\" — AI Research Acceleration Report, 2024\n\nImportantly, this does not replace the human-centered ethos of design thinking. AI handles the operational overhead — scheduling, transcription, initial theme extraction — so researchers can focus on the interpretive work that requires human judgment: synthesizing empathy maps, forming HMW statements, and building the understanding that drives design.\n\nKoji's platform supports the full Empathize workflow:\n- **Open-ended questions** for narrative, exploratory conversation\n- **Structured questions** ([scale, single_choice, ranking](/docs/structured-questions-guide)) to quantify the prevalence and intensity of specific findings\n- **Automatic transcript analysis** to extract themes without manual coding\n- **Research reports** that organize findings in the format needed for the Define phase\n\n---\n\n## Designing Your Empathy Research Study\n\n### How Many Participants?\n\nThe Empathize phase is not about statistical significance — it's about insight depth. For early-stage design thinking, **8–15 participants** typically produces sufficient richness for meaningful pattern identification. If you're working with multiple distinct user segments, plan 5–8 participants per segment.\n\nSigns you've reached sufficient insight depth:\n- New interviews are confirming patterns already identified rather than surfacing new ones\n- Your empathy map clusters are well-populated with diverse examples\n- You can write 5–7 HMW statements with confidence\n\n### Who to Research\n\nFor maximum insight, prioritize:\n- **Extreme users:** People who use the product or face the problem at the extremes — the most intensive users and those who barely use it at all. Edge cases illuminate the design space more sharply than average users.\n- **Lead users:** Early adopters and heavy users who have developed sophisticated workarounds and adaptations.\n- **Non-users and churned users:** People who tried and stopped using your product (or a competitor's) often have the most diagnostic perspective on what doesn't work.\n\n### Choosing the Right Research Mode\n\n| Situation | Recommended Approach |\n|---|---|\n| Exploring a new problem space | Open-ended interviews + observation |\n| Validating a defined problem | Structured interviews with quantitative questions |\n| Capturing longitudinal experience | Diary study + follow-up interviews |\n| Reaching large, diverse samples quickly | AI-moderated voice interviews |\n| Behavioral patterns in natural context | Contextual inquiry + shadowing |\n\n---\n\n## Design Thinking vs. Traditional Product Development\n\n| Dimension | Traditional Development | Design Thinking |\n|---|---|---|\n| Starts with | Requirements or features | User needs and context |\n| Research timing | Late validation | Front-loaded empathy |\n| Problem framing | Assumed and fixed | Discovered and refined |\n| Iteration | At end (QA) | Throughout (empathize → test → empathize) |\n| Definition of success | Feature delivery | User problem solved |\n| Risk profile | High (late discovery) | Lower (early learning) |\n\n> \"Design thinking has a bias toward action and empathy toward who you are designing for — to not have a fear of failure.\"\n> — Bernie Roth, Director, Hasso Plattner Institute of Design, Stanford\n\n---\n\n## Getting Started With Empathy Research\n\n**Step 1: Define your research question.** What aspect of your users' experience needs to be understood? Be specific enough to guide research but broad enough to allow surprising discoveries.\n\n**Step 2: Map your assumptions.** Before talking to users, list everything your team believes about them. These are your hypotheses. Empathy research will confirm, refute, or complicate each one.\n\n**Step 3: Recruit diverse participants.** Prioritize range over representativeness at this stage. Include extreme users, edge cases, and people your team doesn't typically talk to.\n\n**Step 4: Run exploratory interviews.** Aim for 45–60 minute sessions. Lead with open-ended narrative questions. Follow the energy — when something interesting surfaces, go deeper.\n\n**Step 5: Build empathy maps.** For each participant and then collectively across all participants. Look for tensions between what users say, think, do, and feel.\n\n**Step 6: Generate HMW statements.** Transform the 3–5 most significant insights into design challenges using the \"How Might We\" format. These feed directly into Ideation.\n\n**Step 7: Validate at scale.** Use AI-moderated interviews to test whether patterns identified in qualitative research hold across a larger, more diverse sample before committing to a design direction.\n\n---\n\n## Related Resources\n\n- [Structured Questions in AI Interviews](/docs/structured-questions-guide) — add scale, ranking, and choice questions to empathy interviews to quantify insights\n- [Empathy Map Guide](/docs/empathy-map-guide) — deep dive on building and facilitating empathy mapping workshops\n- [Generative Research Guide](/docs/generative-research-guide) — the research methods that surface unknown user needs\n- [How to Write User Interview Questions That Surface Real Insights](/docs/user-interview-questions) — question design for deep empathy work\n- [Qualitative Research Software](/docs/qualitative-research-software) — tools to support the design thinking research stack\n- [Prototype Testing and Concept Validation](/docs/prototype-testing-concept-validation) — how the Test phase feeds back into the Empathize phase\n\n## Further reading on the blog\n\n- [B2B Customer Research: The Complete Guide for Product Teams (2026)](/blog/b2b-customer-research-guide-2026) — B2B customer research is harder than B2C — you are navigating buying groups of 10+ stakeholders, gatekeepers, and enterprise procurement cyc\n- [User Research: The Complete Guide to Understanding Your Users (2025)](/blog/user-research-the-complete-guide-to-understanding-your-users-2025) — Learn what user research is, why it matters, and how to conduct it effectively. Discover how AI tools like Koji are transforming the researc\n- [Best AI Market Research Tools in 2026: The Complete Buyer's Guide](/blog/ai-market-research-tools-2026) — AI has fundamentally changed market research. This guide compares the leading AI market research platforms—from AI-native interview tools li\n\n<!-- further-reading:blog -->\n","category":"Research Methods","lastModified":"2026-05-13T00:26:36.807295+00:00","metaTitle":"Design Thinking Research: The Complete Guide to the Empathize Phase","metaDescription":"Master design thinking's Empathize phase with proven user research techniques. Learn empathy mapping, contextual inquiry, How Might We statements, and how AI interviews scale human-centered design research. Includes statistics showing 211-301% ROI from design thinking.","keywords":["design thinking research","empathize phase","design thinking process","empathy mapping","human-centered design","design thinking user research","how might we statements","design thinking empathize"],"aiSummary":"Design thinking is a 5-stage human-centered framework (Empathize, Define, Ideate, Prototype, Test) where research-driven empathy underpins all design decisions. Companies using design thinking outperformed the S&P 500 by 211% over 10 years. The Empathize phase uses user interviews, contextual inquiry, observation, and empathy maps to discover real user needs. Insights are synthesized into \"How Might We\" statements that drive ideation. AI-moderated interview platforms can reduce empathy research time by up to 44% while scaling to hundreds of participants.","aiPrerequisites":["Basic understanding of product development or UX design"],"aiLearningOutcomes":["Understand the 5 stages of design thinking and their purpose","Run empathy interviews that surface behavioral insights","Build empathy maps that synthesize research findings","Write \"How Might We\" statements from research insights","Scale empathy research with AI-moderated interviews"],"aiDifficulty":"intermediate","aiEstimatedTime":"16 minutes"}],"pagination":{"total":1,"returned":1,"offset":0}}