We have more decisions than ever, and fewer researchers to help us answer them

Last week Lenny Rachitsky published his late-2026 read of the tech job market. There is cause to celebrate! There is a 75% increase in product management jobs. There is also cause to worry. In the past year, US tech companies hired about 1,800 user researchers. That’s down from 5300 in 2022. This is a consistent trend in the past few years. Sadly, it’s not a blip but a trend. Our quick Google/Claude search told us that there are 108 open user-research roles at tech companies worldwide right now. Meta and Google made roughly one in eight of the US research hires last year, which means that outside a handful of very large companies, the dedicated researcher is becoming somebody your bosses worked with but you are not likely to be so lucky.

Here’s the punchline: companies are hiring the people whose job is to decide (product managers), and they’re cutting the people whose job was to get the evidence needed to make those decisions.

The questions are still there though

We are not happy about this trend at all. Our position for years has been that while research certainly is a skill that can be democratized to an extent, dedicated researchers and designers make product managers better at their decisions and jobs. That said, the questions PM’s need to answer still remain and that data still needs to be collected.

Think about what a user researcher actually does inside a product team. Forget the deliverables they created and focus on the questions they asked and the data they provided. Someone in a planning meeting says "customers will love this" and the researcher is the person in the room who asks, "which customers, doing what, and how do we know?" That question is still necessary even if the researcher has been laid off. It’s just that now that question gets answered by a PM who has a roadmap to defend and not enough data to do so. Or, worse, it gets answered by an AI tool with a set of synthetic users who, as we wrote last week, always say yes.

The function was retired as if it were a cost center that provided negligible value. But research was never the reports it created. It was the evidence inside the reports that fed into better decision making. When you cut the function, you don’t cut the decisions. You cut the evidence, and the decisions get worse.

Research can't go away, but it has to change

Nielsen Norman Group published a piece on Thursday with a line we agree with completely: research can't go away, but it needs to change. The argument is that AI coding agents have made building and changing software so cheap that teams now ship first and learn later. A release tells you what happened but not why it happened. A researcher's job needs to evolve so that their work involves deciding what each release has to measure, studying what comes back, and running the open-ended discovery that no amount of instrumentation or automation can replace.

NNg then asks probably the most important question. Who reviews the data that comes back? As staff numbers dwindle the folks who wrote the measurement requirements are no longer there to review what comes back from the market. Those who are left don’t have the time to review it either because their workload has now tripled. This isn’t “adapting your research practice.” It’s abandoning it in the hopes that you’ll just learn from whatever is in production and how it performs. This is an expensive strategy, to say the least.

One of the comments on Jeff’s post about synthetic users last week came from Santiago Bustelo, who has been doing interaction design in Latin America for thirty years. He made the point that if a machine could accurately predict what people will do and say, it could predict the economy, and nobody would be renting it out to test websites for twenty dollars a month. A synthetic user, he wrote, is a simulation of a useful tool, not of a person. We would put it this way: the tool is useful precisely when there is a human researcher deciding what it is for and reading what it produces. When you remove that person, the tool becomes just a faster way to be confidently wrong.

What to do if your company stopped hiring researchers

First thing to try: hire them back. Sadly, we know this is going to be close to impossible right now.

Put the question inside the decision, not inside a report. Every meaningful product decision your team makes should carry its own research question, “What would we have to see to know this is wrong, and who is going to go and look?” In Lean UX terms that is the second half of the hypothesis statement, the "we will know we are wrong if" clause. When research was a department, that clause was part of a study someone was running. Now it has to live as part of the decision itself, with the name of the person who is going to find out written next to it.

Make the researchers you still have multipliers, not producers. If you have one researcher for forty product people, her job is not to run forty studies. It is to write the measurement requirements that go into every release, to teach product managers and designers to run the small, frequent studies themselves, and to build the systems that carry what the company learns into more decisions than she could ever sit in on. That is what the people doing this well mean when they talk about moving from research as a service to research as intelligence.

Name the reviewer. Before a feature ships, a specific human needs to be assigned as “accountable” for that feature. Their job is to ensure that not only does the feature work well for customers upon launch but that it continues to do so into the future. And, if it doesn’t, to bring that evidence back and help the company decide what to do about it.

This is exactly what Josh Seiden and I are going to talk about on Tuesday, Nov 17, 2026 with Jen Cardello, who leads experience design research at Capital One. The session is called From UX Research to UX Intelligence. It is free, it is practical, and if you work in product, design or research in a company that is struggling with fitting research into its new AI reality, it’s for you.

Register here: https://webinar.senseandrespond.co/

If the time does not work, register anyway and we will send you the recording.

Next
Next

You talked to synthetic users but never asked the actual humans