{"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":"2026-05-31T19:08:00.260Z"},"content":[{"type":"documentation","id":"712decf4-64a0-4ac4-a6ec-ad1841b113ff","slug":"attitudinal-vs-behavioral-research","title":"Attitudinal vs. Behavioral Research: What Users Say vs. What They Do","url":"https://www.koji.so/docs/attitudinal-vs-behavioral-research","summary":"Attitudinal research captures what users say (interviews, surveys, focus groups). Behavioral research captures what users do (analytics, usability testing, A/B testing). The say-do gap — where stated intent and actual behavior diverge by 38–65% — is why serious researchers combine both. NNG's 2x2 framework maps all research methods across these two dimensions. AI-powered interview platforms like Koji operate in the attitudinal + qualitative quadrant, enabling 10x the interview volume at a fraction of the cost.","content":"## The Core Problem: What People Say vs. What They Do\n\n65% of consumers say they buy from purpose-driven, sustainable brands. Only 26% actually do so. That 39-percentage-point gap — between stated intent and actual behavior — is not an anomaly. It is one of the most consistent findings in consumer and user research.\n\nAnd it is the reason every serious researcher needs to understand the distinction between **attitudinal research** and **behavioral research**.\n\n> \"What people say, what people do, and what people say they do are entirely different things.\"\n> — Margaret Mead, Anthropologist\n\nThis is the foundational insight of the attitudinal vs. behavioral framework. Neither type of research alone gives you the full picture. Together, they form the most powerful diagnostic loop in UX and product research.\n\n---\n\n## The Framework: NNG's 2×2 Matrix\n\nChristian Rohrer at Nielsen Norman Group formalized the canonical taxonomy in \"When to Use Which User-Experience Research Methods\" — one of the most widely cited articles in UX practice. The framework places every research method on two axes:\n\n1. **Attitudinal vs. Behavioral** — what people *say* versus what people *do*\n2. **Qualitative vs. Quantitative** — understanding the *why* versus measuring the *how many*\n\nThe intersection produces four quadrants, each suited to different research goals:\n\n| | **Attitudinal** | **Behavioral** |\n|---|---|---|\n| **Qualitative** | User interviews, focus groups, diary studies | Usability testing (observation), contextual inquiry, field studies |\n| **Quantitative** | Surveys, NPS, Likert scale studies | A/B testing, analytics, eye tracking, clickstream analysis |\n\nNNG also notes a third dimension — **context of use** — ranging from natural use in the wild to scripted lab scenarios to no product involvement at all. But the 2×2 is the practical starting point for deciding which method to reach for.\n\n---\n\n## What Is Attitudinal Research?\n\nAttitudinal research captures **what users think, feel, prefer, and believe** about a product, service, or concept. It answers the *why* questions: Why do users choose this product? What motivates them? What do they value? What do they perceive as difficult?\n\nData is self-reported — collected through questions, prompts, and conversations. This is its strength and its limitation.\n\n### Core Attitudinal Methods\n\n**In-depth user interviews** are the gold standard for attitudinal research. One-on-one conversations that explore mental models, goals, frustrations, and motivations in depth. 8–15 participants typically reach thematic saturation.\n\n**Surveys and questionnaires** scale attitudinal data collection to hundreds or thousands of participants. Best for measuring satisfaction (NPS, CSAT), stated preferences, and demographic patterns.\n\n**Focus groups** surface shared attitudes and top-of-mind reactions to concepts. Useful for early-stage concept testing and understanding the language users use to describe problems.\n\n**Diary studies** capture self-reported attitudes and experiences in context over time — ideal for longitudinal tracking of how perceptions evolve with extended product use.\n\n**Card sorting and concept testing** reveal users' mental models of information architecture and early-stage concept viability before building.\n\n### Limitations of Attitudinal Research\n\nAttitudinal data is shaped by cognitive biases that researchers must account for:\n\n- **Social desirability bias**: Participants answer in ways they believe are acceptable or impressive rather than truthfully\n- **Recall bias**: Memory of past behavior is reconstructed, not retrieved — and subject to narrative shaping\n- **Hypothetical bias**: People systematically overestimate their future behavior (\"I would definitely use that feature\")\n\nThis is why attitudinal research must be validated with behavioral evidence whenever possible.\n\n---\n\n## What Is Behavioral Research?\n\nBehavioral research captures **what users actually do** when interacting with a product. It observes or logs real actions — clicks, scrolls, task completion, navigation paths, session durations — without asking users to explain themselves.\n\nBehavioral data is objective in a way attitudinal data cannot be. Users cannot \"misremember\" a click. But it tells you *what* happened without telling you *why*.\n\n### Core Behavioral Methods\n\n**Usability testing** (observational) watches users complete specific tasks with a product, revealing friction points, confusion, and failure patterns. Even with 5 users, approximately 85% of major usability issues surface.\n\n**Product analytics** aggregate behavioral data at scale — funnels, drop-off rates, feature adoption, session recordings, and cohort retention. Essential for identifying *where* problems occur across your entire user base.\n\n**A/B testing** compares design variants based on measurable behavioral outcomes — conversion rates, engagement, clicks — with statistical significance.\n\n**Eye tracking and heatmaps** visualize where users look and click on a page, revealing attention patterns that users themselves cannot articulate.\n\n**Clickstream and session recording** provides full playback of individual user sessions — every scroll, hesitation, and error — giving qualitative texture to quantitative patterns.\n\n### Limitations of Behavioral Research\n\nBehavioral data tells you what happened but not why. A 70% drop-off on step 3 of your onboarding flow is a fact. The reason — confusion about the terminology, distrust of the permission request, a competing phone notification — is invisible in behavioral data alone. That explanation requires attitudinal follow-up.\n\n---\n\n## The Say-Do Gap: Why You Need Both\n\nThe **say-do gap** is the empirically documented disparity between what users claim they will do and what they actually do. It is the central argument for mixed-method research.\n\nHard evidence of the say-do gap in practice:\n\n- **38% of US online shoppers do not follow their previously stated behavior** — 4 in 10 people act differently from what they told researchers\n- **65% of consumers say they buy from sustainable brands; only 26% actually do** — a 39-point gap between stated values and purchasing behavior\n- Companies that build product roadmaps from survey data alone routinely ship features with strong stated demand that users abandon in practice\n\nNielsen Norman Group frames it this way: *\"Users often misremember past actions, experience social-desirability bias, and struggle articulating internal experiences. Consequently, what users report often diverges significantly from their actual behavior — making these mismatches valuable sources of design insights.\"*\n\nThe mismatch itself is data. When users say they find a feature important but never use it, that tension reveals a gap between perceived value and realized value — exactly the kind of insight that drives product strategy.\n\n### The Gold-Standard Diagnostic Loop\n\nThe most powerful research pattern is using behavioral data to identify problems, then attitudinal data to explain them:\n\n1. **Analytics flag a drop-off** on a specific step (behavioral)\n2. **Interviews explore why** users abandon at that step (attitudinal)\n3. **Usability testing validates** whether proposed fixes actually improve completion (behavioral)\n4. **Follow-up surveys measure** whether satisfaction improved post-fix (attitudinal)\n\nThis loop — behavioral → attitudinal → behavioral → attitudinal — is how the best product and research teams drive decisions.\n\n---\n\n## When to Use Each: A Decision Framework\n\n### Reach for Attitudinal Research When:\n- You are in **early discovery** and need to understand user goals, mental models, and motivations before building\n- You want to understand **why users make specific choices** or feel certain ways about a product\n- You are **exploring unmet needs** or future product directions\n- You need to measure **perceived usability or satisfaction** (NPS, CSAT, SUS)\n- You are testing **concepts or prototypes** before development investment\n\n### Reach for Behavioral Research When:\n- You need to understand **what users actually do** — navigation patterns, feature adoption, task completion\n- You are **diagnosing specific UX problems** identified in analytics\n- You need **quantitative evidence** to justify design decisions to stakeholders\n- You are **comparing design variants** through A/B testing\n- You want insight at **scale** — thousands of users, not dozens\n\n### Use Both Together When:\n- Behavioral data has surfaced a problem and you need attitudinal data to explain it\n- You are building a **new product from scratch** (attitudinal for discovery, behavioral for validation)\n- You are **running usability testing** (behavioral observation) and want to follow up with questions (attitudinal)\n- You want to check whether stated preferences actually predict usage behavior\n\n---\n\n## The ROI Case for Getting This Right\n\nThe cost of using the wrong research type at the wrong time is significant:\n\n- **Every $1 invested in UX research returns up to $100** in downstream savings — IBM's \"1:10:100 rule\" shows that fixing a problem costs $1 during research, $10 in development, and $100 after launch (Forrester Research)\n- **Companies in the top quartile of design practice** — which includes systematic user research — achieve 32% higher revenue growth and 56% higher total shareholder returns than peers (McKinsey, 2018)\n- **95% of new products fail**. 34% of startups cite insufficient customer understanding as the primary cause — the result of either skipping research entirely or using attitudinal data (stated interest) to validate behavioral adoption questions\n\nThe IBM 1:10:100 rule applies to research method selection too: using attitudinal research to answer behavioral questions (or vice versa) produces expensive misdirection that compounds with every sprint.\n\n---\n\n## AI-Powered Attitudinal Research at Scale\n\nThe most significant practical limitation of attitudinal research has historically been **scale**. A skilled researcher can conduct 4–6 in-depth interviews per week. Most teams complete 5–10 interviews per quarter. NNG's \"5 users\" rule emerged partly because recruiting and moderating more interviews is practically difficult.\n\nAI moderation removes this constraint. Platforms like Koji sit squarely in the **Attitudinal + Qualitative** quadrant — conducting the same type of research as human-moderated interviews — but without the scheduling, moderation, and synthesis bottlenecks.\n\n**What this changes in practice:**\n- Run 50–100 attitudinal interviews in a week instead of a quarter\n- Achieve consistent probing across every participant — no moderator bias, no fatigue-induced shortcuts\n- Synthesize themes, sentiment patterns, and emerging signals automatically across hundreds of conversations\n- Pair with Koji's [structured questions](/docs/structured-questions-guide) — using scale, single_choice, yes_no, and ranking question types — to blend attitudinal depth with quantitative signal in a single study\n\nOne additional advantage: AI interviewers naturally implement the Mom Test principle — asking about past behavior (\"Tell me about the last time you...\") rather than hypothetical future intent (\"Would you use a feature that...\"). This behavioral anchoring within attitudinal questioning directly reduces the say-do gap in collected data.\n\n> \"The AI did not ask leading questions the way human moderators often do. It asked open, curious follow-ups that produced some of the richest responses.\"\n> — Koji research team observation\n\n**AI adoption in research is accelerating:** LLM usage in survey and research contexts grew from 1.6% in 2023 to 59% in 2024. Research teams not using purpose-built AI tools are 4× more likely to lose organizational influence than those who do (Qualtrics, 2026).\n\n---\n\n## Practical Checklist: Quick Reference\n\n**Use attitudinal research when you need to know:**\n- Why users feel a certain way\n- What motivates or blocks adoption\n- Whether a concept resonates before you build it\n- How users describe their problems in their own words\n\n**Use behavioral research when you need to know:**\n- Where users drop off\n- Which features are actually used (vs. valued)\n- Whether a design change improves task completion\n- What patterns emerge across thousands of sessions\n\n**Combine both when you want:**\n- Explanations for behavioral anomalies found in analytics\n- Validated findings (not just stated preferences) from a discovery study\n- The most credible evidence base for a high-stakes design or product decision\n\n---\n\n## Related Resources\n\n- [Structured Questions in AI Interviews](/docs/structured-questions-guide) — blend attitudinal depth with quantitative signal using 6 question types\n- [How to Analyze Qualitative Data](/docs/how-to-analyze-qualitative-data) — turn raw attitudinal data into actionable insights\n- [The Definitive Guide to User Interviews](/docs/user-interview-guide) — master the gold-standard attitudinal method\n- [Qualitative vs. Quantitative Research](/docs/qualitative-vs-quantitative-research) — the full methodology breakdown\n- [AI-Moderated Interviews: How Automated Research Works](/docs/ai-moderated-interviews) — how Koji conducts attitudinal research at scale\n- [How to Analyze Qualitative Data](/docs/how-to-analyze-qualitative-data) — from transcripts to themes\n\n## Further reading on the blog\n\n- [How to Analyze Customer Interview Data: A Complete Guide](/blog/how-to-analyze-customer-interview-data) — You ran the interviews. Now what? Here is a step-by-step process for turning raw transcripts into clear, actionable insights your team will \n- [Surveys vs Interviews: When to Use Each (And When to Use Both)](/blog/survey-vs-interview-when-to-use) — Surveys give you scale. Interviews give you depth. But choosing the wrong method wastes time and produces data you cannot act on. Here is a \n- [AI-Moderated vs Human-Moderated Interviews: Which Should You Choose?](/blog/ai-moderated-vs-human-moderated-interviews) — AI-moderated and human-moderated interviews each have a time and a place. Here is the honest comparison to help you choose the right approac\n\n<!-- further-reading:blog -->\n","category":"Research Methods","lastModified":"2026-05-13T00:25:38.788654+00:00","metaTitle":"Attitudinal vs. Behavioral Research: What Users Say vs. What They Do","metaDescription":"Understand the difference between attitudinal and behavioral research with NNG's 2x2 framework. Learn when to use each method, why the say-do gap matters, and how AI scales attitudinal research.","keywords":["attitudinal research","behavioral research","attitudinal vs behavioral","say do gap","NNG framework","UX research methods","qualitative research","quantitative research","user research methods"],"aiSummary":"Attitudinal research captures what users say (interviews, surveys, focus groups). Behavioral research captures what users do (analytics, usability testing, A/B testing). The say-do gap — where stated intent and actual behavior diverge by 38–65% — is why serious researchers combine both. NNG's 2x2 framework maps all research methods across these two dimensions. AI-powered interview platforms like Koji operate in the attitudinal + qualitative quadrant, enabling 10x the interview volume at a fraction of the cost.","aiPrerequisites":["Basic understanding of UX research methods"],"aiLearningOutcomes":["Understand the NNG 2x2 attitudinal/behavioral framework","Know when to use attitudinal vs behavioral methods","Recognize the say-do gap and why it matters","Apply the diagnostic loop combining both method types","Scale attitudinal research with AI-powered interviews"],"aiDifficulty":"intermediate","aiEstimatedTime":"12 minutes"}],"pagination":{"total":1,"returned":1,"offset":0}}