Ethnographic Research In UX

There's something almost nobody is talking about in AI - but it affects everything from asking ChatGPT for advice to companies deploying AI globally. A fascinating study tested major AI Models - the foundations powering tools millions use daily - against cultural values from 107 countries worldwide. The result? Each one reflected the same assumptions - those of English-speaking, Western European societies. None aligned with how people in Africa, Latin America, or the Middle East actually build trust, show respect, or resolve conflicts. Why does this matter? Imagine you're a global company rolling out AI customer service. Your system learns "best practice": when customers complain about late orders, "apologise briefly, offer a discount, and focus on quick resolution". In Germany, the direct, efficient approach works perfectly. Customer satisfied. But in Japan, that brief apology violates meiwaku - the cultural need to deeply acknowledge when you've caused someone inconvenience. Your "efficient" response feels dismissive and damages customer relationships. And in the UAE, the discount offer backfires completely. It feels like charity rather than respect. One AI system, similar contexts, completely different cultural outcomes. This isn't intentional though - it's inevitable. LLMs absorb embedded patterns about communication from their training data, and most of that data comes from billions of English web pages and content. The result? AI systems that, unless thoughtfully shaped, are blind to the diversity of human interaction. Klarna, the global payments company, made headlines in 2024 when they introduced an AI system that "did the work of 700 customer service reps", handled 2.5 million conversations in 35 languages, and cut response time by 82%. Technical triumph. 14 months later: "Klarna reverses AI strategy and is hiring humans again". Their CEO admitted it had led to "lower quality". Some reports said they'd seen a 20%+ decrease in customer satisfaction. What I think really happened: Klarna optimised for 35 languages while completely missing 35 different ways humans expect to be treated. The challenge? Most companies are focusing on technical integration and completely missing cultural intelligence. We measure response time and cost savings, but never ask, "which human complexities are we overlooking?" The goal isn't neutrality though - that's impossible and undesirable. It's conscious awareness. Understanding that the output from AI models is filtered through a specific cultural lens. For companies building AI strategies, key questions worth asking: * Which cultural assumptions are embedded in our AI systems? * How do we test cultural intelligence alongside technical performance? * Who provides this expertise in our AI teams? The individuals and organisations that develop this conscious awareness will make better decisions, while others unknowingly apply one-size-fits-all approaches to beautifully diverse human contexts.

Your tech solutions might be universal, but business cultures rarely are. For founders expanding globally, understanding cultural nuances can make a world of difference. I've seen so many brilliant construction tech solutions face unexpected challenges internationally not because of product issues, but because of cultural cues that were hiding in plain sight. What works smoothly in your home market frequently encounters unexpected barriers abroad. In our latest Practical Nerds episode, Shubhankar and I explored three cultural patterns we've observed that often create unexpected challenges for founders expanding internationally: 1/ Trust deficit can kill deals in Asia before you realize what happened. Asian markets require relationships BEFORE transactions. That mid-deal silence? It's not disinterest—it's a fundamental lack of trust. When things stall, don't send another "just checking in" email. Request a direct call: "Hey, can we get on a call? I'd just like to hear from you." 2/ Europeans want facts, not hype. Your high-energy American pitch style? It can be "overcompensating" to Europeans. They're engineering-minded—lead with observations, not judgments. And remember: Europeans minimize downside before maximizing upside. Frame your solution as risk mitigation first, opportunity second. 3/ Middle East surprisingly loves American tech but demands in-person presence. Virtual meetings barely register as "meetings" at all. And forget the org chart—decisions flow through specific gatekeepers who might not even appear in formal hierarchies. What seems to work well for many companies in global expansion? Maintaining consistent products and channels while building localized teams who can navigate the nuances of each market's business culture. 👇 Dive deeper into our full analysis of global construction tech expansion below. #ConstructionTech #GlobalExpansion #BusinessCulture

The real threat isn't malware. It's silence. We had the tools. We had the budget. But we still got breached. That line came from a client years ago in Southeast Asia, and it stuck with me. After 15 years in cybersecurity, one lesson stands out clearly across the Asia Pacific: Technology rarely fails first. People do. When you operate across countries like Singapore, Thailand, Malaysia, Indonesia, and the Philippines, you start to see deeper patterns. Not technical ones. Behavioural ones. In some markets, staff members hesitate to report incidents for fear of being blamed. In others, there is a strong culture of hierarchy, which makes it hard for junior employees to challenge suspicious activity. Sometimes, the biggest obstacle is not the attacker outside. It is the silence inside. We often treat cybersecurity as a technical challenge. But adoption is driven by trust, context, and cultural relevance. That is why the same selling approach that works in the US has a low chance of working in Southeast Asia. It is not because people do not care. It is because the framing does not fit the environment. Over time, I have learnt to approach cyber selling like market development: → Start with empathy → Speak the local language, not just linguistically but emotionally and socially → Focus on behaviours, not just knowledge Cybersecurity in Asia is not just about rolling out tools. It's about building a culture where people feel safe speaking up, slowing down, and making informed decisions under pressure. If we want real resilience, we need to stop selling fear and start shaping habits. What cultural or behavioural barriers have you seen when it comes to cyber awareness in your region? I would love to hear your stories. #alvinsratwork ✦ #BusinessTechnologist ✦ #ExecutiveDirector

Most teams aren’t building personalization in generative recommender systems. They’re building an expensive way to re-read user logs. A common pattern I see: → take a user’s months of interaction history → dump it into an LLM context window → expect high-quality personalization This breaks quickly. Latency blows up, costs scale with tokens, signal quality actually drops, and the model still misses what matters. The "Lost in the Middle" effect is very real in long-context personalization. So teams try: "Let’s just pass the last N interactions." This is the truncation trap because user intent isn’t just recent, it’s layered. → long-term preferences are stable and slow-changing → short-term intent is session-specific and fast-changing. In generative recsys, this is not a context window problem; it’s a memory routing problem. The architecture that works separates memory into two systems: [1] Long-term memory (offline layer) Compress history into a dense user summary: → Run an async summarization pipeline (e.g., nightly) → Maintain a dense user profile capturing stable preferences → Keep it small (on the order of a few hundred tokens) [2] Short-term memory (online retrieval layer) For immediate intent, use retrieval, not raw history. → Store interactions in a vector database (e.g., HNSW index) → At inference, retrieve only the top-K relevant items → Keep it tightly scoped to the current task Then at generation time, combine both: → Long-term summary (global preference) → Retrieved short-term context (local intent) The result is a system that preserves personalization quality, reduces latency significantly, and cuts token usage and cost by an order of magnitude. The key idea: what you choose not to send to the model is just as important as what you include. This is Part 5 of my Generative RecSys series. Next up: I'll break down some of the evaluation aspects for these complex systems. Curious: Where are you drawing the line between latency, cost, and context quality in your Generative RecSys stack?

While it can be easily believed that customers are the ultimate experts about their own needs, there are ways to gain insights and knowledge that customers may not be aware of or able to articulate directly. While customers are the ultimate source of truth about their needs, product managers can complement this knowledge by employing a combination of research, data analysis, and empathetic understanding to gain a more comprehensive understanding of customer needs and expectations. The goal is not to know more than customers but to use various tools and methods to gain insights that can lead to building better products and delivering exceptional user experiences. ➡️ User Research: Conducting thorough user research, such as interviews, surveys, and observational studies, can reveal underlying needs and pain points that customers may not have fully recognized or articulated. By learning from many users, we gain holistic insights and deeper insights into their motivations and behaviors. ➡️ Data Analysis: Analyzing user data, including behavioral data and usage patterns, can provide valuable insights into customer preferences and pain points. By identifying trends and patterns in the data, product managers can make informed decisions about what features or improvements are most likely to address customer needs effectively. ➡️ Contextual Inquiry: Observing customers in their real-life environment while using the product can uncover valuable insights into their needs and challenges. Contextual inquiry helps product managers understand the context in which customers use the product and how it fits into their daily lives. ➡️ Competitor Analysis: By studying competitors and their products, product managers can identify gaps in the market and potential unmet needs that customers may not even be aware of. Understanding what competitors offer can inspire product improvements and innovation. ➡️ Surfacing Implicit Needs: Sometimes, customers may not be able to express their needs explicitly, but through careful analysis and empathetic understanding, product managers can infer these implicit needs. This requires the ability to interpret feedback, observe behaviors, and understand the context in which customers use the product. ➡️ Iterative Prototyping and Testing: Continuously iterating and testing product prototypes with users allows product managers to gather feedback and refine the product based on real-world usage. Through this iterative process, product managers can uncover deeper customer needs and iteratively improve the product to meet those needs effectively. ➡️ Expertise in the Domain: Product managers, industry thought leaders, academic researchers, and others with deep domain knowledge and expertise can anticipate customer needs based on industry trends, best practices, and a comprehensive understanding of the market. #productinnovation #discovery #productmanagement #productleadership

If Behaviors Drive Culture... what drives behavior then? We all know that culture shapes what feels expected, what gets rewarded, and what people learn to avoid in an organization, but how do we get to the root cause of it all? If we want to change or evolve culture, we can’t just stop at naming the behaviors we’d like to see more of, we also have to understand what’s making those behaviors hard or misaligned with how things currently work. I took this simple culture map, and added a few examples to help illustrate how it works. This isn't a template to copy, because of course it is incomplete, but I like to show things so they are visual and easy to understand. This mapping or canvas becomes useful, because it helps your team see what’s underneath the behavior... think beliefs norms, expectations, structures, and feedback loops that quietly shape what people do, even when no one is looking. This is the part that most people get wrong... because they go from what’s visible to what they think is a solution, without understanding what kind of barrier they’re dealing with. This is where real behavioral change needs to become operational! You need a way to classify what’s going on - you need models and frameworks to help answer: • Is it a lack of skill or knowledge? • A belief that’s outdated but persistent? • A social norm no one questions? • A structural constraint that makes the behavior impractical? •An identity issue, where doing the new behavior would mean not being seen as a “good colleague” or “strong leader”? If you can’t distinguish these nuances, then any solution you design is essentially a guess because you are working with intuition and not evidence. This is the most important part to get right and where most of my clients ask me to work with them on, because In every project we run, whether it’s cultural alignment, policy redesign, or transformation strategy, we build this kind of behavioral mapping in from the start. 1) We identify the behaviors that matter. 2) We map what’s reinforcing or blocking them. 3) We use models like COM-B, TDF, and BNOS to analyze the patterns. 4) We match techniques to the actual barrier so the solution fits the mechanism, not just the symptoms. Most of my careers, I used to go straight from behavior to idea, without doing the work in the middle, but once I learned how to build this kind of analysis into the process, things started working more consistently and more systemically. In the end, if you want to change behavior, you have to understand what holds it in place and If you want to evolve culture, that’s where you start. What’s quietly holding your current behaviors in place, even when everyone says they want change? (Canvas adapted from - Dave Grey/Strategyzer)

💡 Mapping user research techniques to levels of knowledge about users When doing user research, it's important to choose the right methods and tools to uncover valuable insights about user behavior. It's possible to identify 3 layers of user behavior, feelings, and thoughts: 1️⃣ Surface level - Say & Think This level captures what users say in conversations, interviews, or surveys and what they think about a product, feature, or experience. It reflects their stated opinions, thoughts, and intentions. Example: "I prefer simple products" or "I think this app is easy to use." Methods: Interviews, Questionnaires. These methods capture stated thoughts and opinions. However, insights may be influenced by social norms or biases. 2️⃣ Mid-level - Do & Use This level reflects what users actually do when interacting with a product or service. It emphasizes actions, usage patterns, and observed behaviors, revealing insights that may differ from what users say. Example: Users may claim they enjoy customizing app settings, but data shows they rarely change default options. Methods: Usability Testing, Observation. Observation helps to reveal gaps between what people say and what they actually do. 3️⃣ Deep level - Know, Feel and Dream This level uncovers deep motivations, emotions, desires, and aspirations that users may not be consciously aware of or may struggle to articulate. It also includes tacit knowledge—things people know intuitively but find hard to express. Example: A user might not realize that their preference for a minimalist design comes from the information overload of a current design. Methods: Probes (e.g., participatory design, diary studies). Insights collected using these methods will uncover implicit and emotional drivers influencing behavior. 📕 Practical recommendations for mapping ✅ Triangulate insights by using multiple methods. What people say (interviews/surveys) may differ from what they do (observations) and feel. That's why it's essential to interpret these results in context. For example, start with interviews to learn what users say. Follow up with usability testing to observe real behavior. Use probes for long-term or emotional insights. ✅ Align research with business goals. For product improvements, focus on usability testing to catch interaction issues. For innovation, use probes to generate new ideas from user insights. ✅ Practice iterative learning. Apply surface techniques (like surveys) early to refine assumptions and guide more in-depth research later. Use deep techniques (like probes) for strategic decisions and to foster innovation in long-term projects. 🖼️ UX Research methods by Maze #ux #uxresearch #design #productdesign #uxdesign #ui #uidesign

Qualitative research in UX is not just about reading quotes. It is a structured process that reveals how people think, feel, and act in context. Yet many teams rely on surface-level summaries or default to a single method, missing the analytical depth qualitative approaches offer. Thematic analysis identifies recurring patterns and organizes them into themes. It is widely used and works well across interviews, but vague or redundant themes can weaken insights. Grounded theory builds explanations directly from data through iterative coding. It is ideal for understanding processes like trust formation but requires careful comparisons to avoid premature theories. Content analysis quantifies elements in the data. It offers structure and cross-user comparison, though it can miss underlying meaning. Discourse analysis looks at how language expresses power, identity, and norms. It works well for analyzing conflict or organizational speech but must be contextualized to avoid overreach. Narrative analysis examines how stories are told, capturing emotional tone and sequence. It highlights how people see themselves but should not be reduced to fragments. Interpretative phenomenological analysis focuses on how individuals make meaning. It reveals deep beliefs or emotions but demands layered, reflective reading. Bayesian qualitative reasoning applies logic to assess how well each explanation fits the data. It works well with small or complex samples and encourages updating interpretations based on new evidence. Ethnography studies users in real environments. It uncovers behaviors missed in interviews but requires deep field engagement. Framework analysis organizes themes across cases using a matrix. It supports comparison but can limit unexpected findings if used too rigidly. Computational qualitative analysis uses AI tools to code and group data at scale. It is helpful for large datasets but requires review to preserve nuance. Epistemic network analysis maps how ideas connect across time. It captures conceptual flow but still requires interpretation. Reflexive thematic analysis builds on thematic coding with self-awareness of the researcher's lens. It accepts subjectivity and tracks how insights evolve. Mixed methods meta-synthesis combines qualitative and quantitative findings to build a broader picture. It must balance both approaches carefully to retain depth.

Two quiet team members. Two different reasons. Vietnamese culture or who they are? If you're guessing, your leadership shows it. You can't tell the difference. That's the problem. You read that "Vietnamese people value harmony." So you assume Minh won't disagree because he's Vietnamese. That's not cultural intelligence. That's generalization. Culture gives you the pattern. The person shows you where they stand on it. Yes, in Vietnam, direct disagreement with a senior leader is often avoided. That's the cultural pattern. But Minh: he's more direct than most. That's the person. If you assume he'll never push back, you'll miss his subtle signals completely. You'll think he agrees when he's actually disengaging. You need both layers: The cultural lens tells you what's common. The individual understanding tells you what's true for this person. That's cross-cultural leadership. Without the cultural lens, you misread everything. Without the individual lens, you stereotype everyone. Why this matters more than you think: When you get it wrong, your leadership doesn't land. Your feedback feels like an attack. Your delegation feels like distrust. Your praise feels empty. Your team don't bring you ideas. They don't take initiative. They smile, nod, and wait for you to leave. You're leading, but nothing's moving. That silence you're hearing: It's not agreement. It's distance. The expat leaders I work with master both layers. They understand the deep cultural patterns that shape Vietnamese workplaces. Then they see each person clearly within that context. They stop guessing. They stop assuming. They influence ad have impact. If this hits close to home, let’s talk.

Unpopular opinion: AI cultural bias might be your biggest overlooked business risk in 2024. Everyone talks about AI security and data privacy. Few talk about AI worldview. Here's what most executives miss: You think you're buying intelligence. You're actually buying culture. Harvard just proved ChatGPT thinks like Western Europe, not like your global customers. This shows up in three ways that hurt your bottom line: Customer Experience Failures ↳ AI chatbots responding inappropriately to non-Western communication styles ↳ Product recommendations that miss cultural preferences ↳ Marketing messages that don't resonate outside Western markets Hiring and HR Mistakes  ↳ Resume screening that favors Western communication patterns ↳ Interview assessments biased toward individualistic responses ↳ Performance evaluations that miss collectivist work styles Strategic Decision Gaps ↳ Market analysis that misses non-Western consumer behavior ↳ Risk assessments based on Western business norms ↳ Innovation ideas that don't translate globally The fix isn't complicated. It's intentional. Start with one question: "Whose perspective is missing from this AI output?" Test your AI tools with diverse user groups. Build cultural review checkpoints into AI workflows. Partner with regional teams to validate AI recommendations. 62% of companies using biased AI report decreased revenue. 61% have lost customers due to cultural misalignment. The companies winning global markets aren't just using AI. They're using culturally aware AI. Audit one AI system this week for cultural blind spots. What did you discover? ♻️ Share this to help executives build more inclusive AI strategies. ➕ Follow me for more insights on global AI implementation.