More than 40% of companies still don’t talk to their end users during development. That figure comes from a McKinsey study that tracked design practices at 300 publicly listed companies over five years. Yet, while remote user testing has made research more accessible than ever with global panels, overnight unmoderated sessions, AI-based analysis and results that come back in days, if not hours, most product teams still skip it before they ship.
A recent interview with Laura Klein, UX designer, Principal Experience Specialist at Nielsen Norman Group, and author of Build Better Products and UX for Lean Startups, makes the cost of that gap concrete. Klein argues that skipping user research remains one of the most expensive decisions a product team can make. In her words,“you’re not moving faster. You’re just shipping the wrong things faster.”
What Skipping UX Research Actually Costs
The failures that come from ignoring user research rarely make it on public record. Klein, who has spent 30 years watching teams make this mistake, has seen what the consequences look like up close.
In one case she witnessed directly, a team overrode research warnings about a product users didn’t understand and were unlikely to keep using. The result was roughly a billion dollars in lost revenue. In another, a team spent two years building something they eventually didn’t ship, pulling budget away from features users had been asking for throughout.
The financial data supports the pattern broadly. The same McKinsey research found that companies in the top quartile for design outpaced their industry peers by 32 percentage points in revenue growth over five years. The separating factor wasn’t budget or headcount, but consistent discipline around listening to users before, during, and after building.
Why the ‘Speed’ Argument Is Backwards
The case for skipping research is often framed as a trade-off: spend time on research now, or spend time on the product now. Klein’s argument is that this framing misrepresents where the time actually goes.
“You’re not moving faster. You’re just shipping the wrong things faster,” asserts Klein.
Let’s consider what rework actually costs organizations. A feature that takes two months to build but doesn’t address a real user need is a complete loss. Research that prevents that build, or redirects it before a single line of code is written, is a gain even if it takes several weeks. The question isn’t whether research takes time but whether the rework will take more. More resources, time, cost, re-work or loss.
Klein’s point is that user research accelerates delivery precisely because it prevents teams from building the wrong thing. She is also deliberate about what counts as ‘useful.’ She isn’t prescribing year-long studies or enterprise-scale discovery programs. A round of usability sessions, a targeted discovery interview series, a quick unmoderated test on a prototype: any focused research activity narrows the solution space and reduces the risk of building something nobody ends up using. “Almost any kind of research will actually end up speeding you up.” The goal is to pick the method that addresses the riskiest unknown and not the most exhaustive one.
“There’s no point doing a million usability studies if the real risk is that we aren’t doing enough discovery research.”
Laura Klein, Principal Experience Specialist, Nielsen Norman Group
Quantitative vs. Qualitative Research: What the Difference Actually Means
Most product teams default to analytics as their primary signal. Dashboards are always on, data is real-time, and numbers translate easily into stakeholder conversations. The limitation here is that metrics show you where something went wrong, but not why.
Klein recalls a meeting where a low-performing metric became the subject of intense debate. Senior stakeholders and 15 minutes of proposed fixes: discounts, UI adjustments, feature changes. Eventually Klein asked the question nobody had thought to ask: why was the metric low in the first place? The room had no answer.
“Quant tells you what’s going on in your product. Qualitative tells you why it’s going on. That’s it. That’s the whole thing.”
The distinction matters because the cause determines the fix. A checkout abandonment rate that is high because the flow is confusing requires a UX intervention. If the same metric is high because a trusted payment method is missing, that is a different problem entirely. And if the abandonment is happening because a form field errors on certain devices, neither of those solutions helps at all. Running analytics-driven guesses at any of these scenarios costs time and tends to introduce new problems. This is where qualitative research eliminates the guesswork.
On the flip side, quantitative data identifies where users drop off. Qualitative research, run through moderated or unmoderated sessions, surfaces the human reasoning behind those numbers and hence why both research methods should come into play.
Where Teams Actually Lose Time
The most legitimate objection to research isn’t about its value, so much as the time required to make sense of what’s been collected. Running sessions are generally perceived as fast, but the bottleneck has always been what comes after: watching recordings, tagging moments, and synthesizing findings into something a product team can actually use. And now with artificial intelligence, the “time to insight,” from human participant user testing sessions to AI produced analysis and reporting, has shrunk dramatically.
One practical lever is matching the method to the question. A team that wants to know whether users can navigate a redesigned information architecture does not need moderated interviews. In this case, tree testing, where participants click through a text-only version of the navigation to find specific content, can be set up and completed with a recruited panel in a matter of days or even hours, and the results are quantitative enough to act on directly. A team validating whether a new onboarding flow makes sense to first-time users can run unmoderated usability sessions where participants attempt the flow and narrate their reasoning as they go, without a researcher present. For deeper discovery work, where teams need to understand motivations or mental models rather than usability behavior, moderated sessions are the appropriate method.
The other lever is analysis speed. Historically, synthesizing qualitative sessions has required researchers to watch recordings, code observations, and build thematic summaries manually. Today’s AI-powered tools for UX research remove this altogether.
Take AI Insights, Userlytics’ new AI assistant inside the platform. Once sessions are completed, AI Insights generates an automatic summary of each session and allows researchers to interactively query the system and ask questions across the full study. Every answer cites the specific session, timestamp, and annotation it drew from, so findings are grounded and verifiable. What previously required two days of manual review can become a focused two-hour conversation with the data.
Userlytics’ new AI-powered tools to accelerate UX research
AI Annotations
The moment transcripts are ready, AI Annotations scans each session for key moments, sentiment patterns, and usability friction, and generates a highlight reel and a chronological list of timestamped, editable annotation cards. Every AI-generated card is flagged so researchers can refine, delete, or add their own. What used to take half a day of session review fits inside a focused pass. Learn more about AI Annotations
AI Insights
Once study sessions are complete, AI Insights provides an initial summary and lets researchers query the system and ask questions across the entire study in line with their overall objectives. Every answer cites the specific session, timestamp, and annotation it drew from, so findings are grounded and verifiable. Teams that previously spent two days in manual synthesis can move through the same work in two hours. Learn more about AI Insights
AI Chart Summaries
On the quantitative side, AI Chart Summaries automatically interprets metrics data and surfaces the most significant patterns without requiring a separate analysis. It answers the “what” faster so researchers can get to the “why” sooner. Learn more about AI Chart Summaries
AI Transcript Translations
Researchers can now translate global studies into their preferred language, choosing from 15 of the most common languages used in our platform. The full transcript updates in seconds and is synced to video playback, with every team member seeing the translation instantly. Learn more about AI Transcript Translations
It’s important to note that the point is not that AI replaces the researcher’s judgment. Klein’s caution on this is clear: outsourcing the thinking to AI is where teams get into trouble. But using AI to surface patterns across a large session set, and then having a researcher review and validate those patterns, is a meaningful acceleration of the synthesis step without sacrificing the interpretive rigor that makes research actionable.
One Habit to Start
For teams trying to move away from shipping-first reflexes, Klein’s starting recommendation is specific: run post-mortems on everything that ships. “Postmortem everything. How did a thing do, and why did it do that way, and how can we do better in the future?”
Post-mortems force attention backward before rushing forward. They surface the assumptions that proved wrong, the research that wasn’t done, and the signals that were set aside under deadline pressure. They also build institutional memory, making it harder to repeat the same mistakes and easier to notice the pattern when skipped research shows up consistently in the retrospective.
The habit is small in execution and significant in what it changes: it shifts the team’s attention from whether a feature shipped to whether it worked.
UX Research for the Win
Skipping proper UX research to save time almost always creates more work downstream. Additionally, a feature rebuilt because it missed the mark costs far more than the research that would have caught the problem earlier. The teams that move fastest and consistently, are the ones that know before they build what they are trying to solve and whether their solution actually addresses it.
