Samantha Shorey

New Article: 20 Cognitive Biases in UX Research This week’s post is a practical guide to the biases that sneak into study design, synthesis, and decision-making. Each one comes with a plain-language definition, a real-world example, and an antidote you can use immediately. Enjoy! https://lnkd.in/eBMfrh9z

60

4 Comments

Why I don’t ask people to think aloud in evaluative research 💭 The ‘think aloud’ method has been a staple of usability testing for decades. But in my experience, the way a lot of researchers use it can introduce some issues: 🚫 Some participants take it too literally, narrating every tiny action. This isn’t how they’d behave at home. 🚫 It makes an already unnatural setting feel even less realistic. 🚫 Once someone starts overanalysing, it’s very hard to bring them back. Instead: ✅ Most participants will talk naturally if you just observe them. ✅ If they’re quiet, you can nudge with simple prompts like “what are you thinking?” I’ve found this leads to more authentic behaviour and better insights than briefing people to ‘think aloud’ from the start. 📨 I cover this in more detail in this week's Desk Notes, my free newsletter. Link in the comments.

60

15 Comments

In current mainstream design practices, chiefs manage strategies to support the business model, design leaders manage the design process, and senior/junior designers deliver a product (or service, environment, system, etc.). In this context, design leaders and designers have to prioritize identifying profitable biases to deliver a product that captures value. However, the generation of value for people is often limited, and they have to endure a product that often deceives them. In a desirable mainstream design practice, design leaders and designers are more strategic and manage up by questioning and influencing their managers. Designers would be more effective at gathering actionable human insights, and people would become collaborators. In this view, people desire products that satisfy them.

166

28 Comments

I’ve loved seeing the energy, referrals, and questions coming in for the recent UXR II opportunity. One question kept popping up: “How do I stand out when I apply?” Here’s my #1 tip for applications: Craft your resume as a story. Don’t just list tasks—show the tangible, quantified impact you have made. Highlight your growth mindset along the way. That shift—impact over activity—helps hiring teams see the value you bring. And because the process doesn’t stop at “Apply,” I also pulled together some interview tips—from the classic STAR method to leaning into your unique strengths and showing AI fluency. 👉 https://lnkd.in/gb64GrCP If you’re on the fence about applying, I encourage you to do so! Diverse paths make stronger products, and yes—the job hunt can feel like a numbers game, but every application is a chance to tell your story.

67

2 Comments

If you are working in UX research, you have likely dealt with count data - metrics that capture how many times something happens, like sessions per week, clicks on a button, or errors in a session. These numbers are always whole numbers and often skew toward lower values. Analyzing them well requires more than just descriptive stats; it means picking the right model for the data’s behavior so your conclusions hold up when product decisions are made. Poisson regression is the starting point for count data. It assumes that the mean and variance are equal, which works when your data is not overly spread out and zeros are expected but not dominant. In practice, though, UX data often has overdispersion - variance greater than the mean - because a small group of users is far more active than the rest. You can detect this by comparing mean and variance or checking a dispersion statistic, and if it’s present, you’ll need a more flexible model. Negative Binomial regression extends Poisson by adding a parameter that lets variance exceed the mean, making it better for data with highly variable user behavior. If your dataset contains more zeros than Poisson or NB would predict, you may need a zero-inflated model. These split users into those who never take the action (structural zeros) and those who might but sometimes don’t (random zeros), letting you account for both processes. When the action happens in two clear steps - first deciding to participate, then deciding how much to do - a hurdle model is more appropriate. It models the binary decision first and then the positive counts, ignoring zeros in the second stage. When your counts are repeated for the same users, Generalized Linear Mixed Models (GLMMs) can add random effects to account for individual differences. If you care more about average effects across the population, Generalized Estimating Equations (GEEs) provide robust estimates without focusing on individual patterns. Fixed Effects Poisson models are useful when you want to compare before-and-after patterns within the same users, controlling for unchanging individual traits without modeling them explicitly. If your counts are collected over time, Count Time Series models like Poisson ARIMA or INGARCH help you capture trends and seasonality. And if you want to understand causal impact - like the effect of a new onboarding flow - Difference-in-Differences compares changes in your treatment group to a control group, isolating the impact of your intervention. Metrics like Incident Rate Ratios (IRRs) and predicted counts help stakeholders see the practical effect of a change, bridging the gap between complex models and decision-ready insights.

98

4 Comments

The great thing about developers leveraging agentic coding tools, is that a lot of the same design best-practices for communicating with humans, are also really helpful context for LLMs. I hope this article is a helpful starting point for designers to set developers up for success, especially those using Figma's Dev Mode MCP server. https://lnkd.in/gNi92tmH

68

Most UX portfolios look the same. Healthcare UX portfolios need to be different. When you're interviewing for healthcare roles, hiring managers aren't just looking for clean interfaces and user flows. They're looking for evidence you understand regulatory constraints, patient safety implications, and clinical workflows. That's a skills gap most UX designers don't know they have. Here's what changes the equation: building something real in healthcare, not just redesigning what already exists. When you can show a working healthcare app you've built, researched, and tested, you're demonstrating capability that goes beyond mockups. That's what this course is designed to do. Over three weeks, you'll build a functional AI-powered healthcare app using no-code tools. You'll learn HIPAA compliance, accessibility standards, and ethical research practices specific to healthcare. Most importantly, you'll finish with portfolio work that shows you can navigate the unique constraints of this field. The course runs on Saturdays with a 2-4 hour weekly commitment. You'll work with a cohort of other professionals making the same transition, which means you're building your network while you're building your skills. If you've been stuck between wanting to specialize in healthcare UX and not knowing how to break in, this gives you a concrete path forward. Next cohorts start October 25 and November 22. Learn more: https://lnkd.in/gaNDBRTc Scholarship info: https://lnkd.in/gyM3Shqi Book a chat: HXRlabs.com

67

There are two main reasons senior design career growth stalls out. 1️⃣ First, they think the more senior they get that bigger and higher impact projects will be handed to them. The reality is that Super ICs often have to create their own scope and impact. You have to increase the scope of existing work, you have to find new opportunities that others don’t see. At the Staff+ level, it’s rare that work just falls in your lap. 2️⃣ Second, they think that the work speaks for itself; that if the craft is strong and they have air-tight rationale that the impact will land. The reality is that at the Staff+ level, influence is the job: storytelling, framing, building coalitions, getting buy-in, crafting a business case, etc. If you can’t do that, your work won’t scale. What other mistakes do you see ICs make as they try to level up?

53

7 Comments

Met with a big UX Research team this week that's working hard to evolve their product in a market disrupted by GenAI summaries. These researchers have a ton of insights in the "old world," and they're working to influence the direction of their "new world." We talked about using some "Stop Wasting Research" book ideas to process their existing insights into useful inputs for product people working to envision adaptations and pivots. When there's a disruption in the market, it's so easy to discount everything that's been learned -- until you really look at that existing work and realize that it can be made extremely useful for today's challenges...

59

5 Comments