Ask a product team why they shipped something without research and you will almost never hear "we didn't think it was valuable." You will hear about the deadline.
That distinction matters more than it looks, because most attempts to fix this problem treat it as a belief problem. More evangelism, more readouts, more insistence that research pays for itself. The teams being evangelized to already agree. They are not skipping research because they doubt it works. They are skipping it because of when it arrives.
The math nobody says out loud
Here is the sequence for a traditional study. Write the brief. Recruit. Wait for the panel to fill. Schedule around eight people's calendars. Run the sessions. Watch the recordings. Synthesize. Write it up. Present it.
Best case, if nothing goes wrong and nobody is on holiday, that is two to three weeks. Realistically it is four to six.
Now look at the decision. It came up on Monday. Engineering has capacity Thursday. The PM needs an answer by Wednesday to shape the sprint. Nobody in that meeting is anti-research. They are looking at a calendar.
Research does not lose that argument on merit. It loses on arrival time. And when an instrument is routinely too slow for the decisions it is supposed to inform, teams do not keep it and wait. They stop reaching for it, and the habit of not reaching for it is what actually kills the practice.
What the barriers actually are
The reported blockers are consistent across surveys, and they are almost all logistics rather than conviction.
Time and bandwidth constraints affect around 63 percent of product and research teams, and roughly 70 percent at enterprise scale. Recruitment is the other perennial: finding the right participants, scheduling them, handling incentives and consent. Teams describe the logistics of a single study taking weeks to arrange before anyone has learned anything.
Nielsen Norman Group has written about the organizational version of this for years, and their diagnosis lands in the same place. The blocker is rarely that leadership thinks research is useless. It is that research is not set up to move at the speed the organization makes decisions.
Then there is access. For plenty of teams the users are genuinely hard to reach: enterprise buyers under NDA, clinicians, people in regulated workflows, or a product that has not launched and therefore has no users at all. No amount of process improvement solves "we cannot get to them."
The staffing arithmetic makes it worse
Even a well-run research function is thin. Nielsen Norman Group's ratio work puts the typical staffing at roughly one dedicated researcher per hundred developers.
Sit with that. A hundred engineers generate a lot of decisions. One researcher cannot be in the room for most of them, and no reasonable amount of prioritization fixes an order-of-magnitude gap. The researcher's job quietly becomes triage: pick the few studies with the best chance of mattering and accept that everything else ships on instinct.
The organization then experiences this as "research is a bottleneck," which is both true and a wildly unfair description of what is happening.
Democratization helps, and creates a new problem
The common response is democratization: train PMs and designers to run their own lightweight studies, with the researcher moving into an enablement role. In 2026 this is close to standard practice, accelerated by AI tooling that compresses transcription, synthesis and clustering.
It genuinely helps with volume. It also creates a quality problem nobody had before, because research skill is not evenly distributed and a badly run study is worse than none. A leading question produces a confident wrong answer, and confident wrong answers get shipped.
So the researcher's role shifts again, from doing research to being accountable for research quality across a group of people who do not report to them and were not trained for it. That is a harder job than the one they had, and most organizations gave them no additional support to do it.
The reframe worth making
Stop asking "how do we get more research done" and start asking "what is the cheapest evidence that beats a guess for this decision."
Not every decision deserves a recruited study, and pretending otherwise is what makes research feel unaffordable. A useful triage:
Reversible and low stakes. Ship it and watch. The instrumentation you already have is the research. Spending two weeks studying a change you can revert in an afternoon is a bad trade.
Consequential and novel. This is what recruited research is for. The decision is hard to undo, you have no prior, and the cost of being wrong is high. Protect this budget aggressively, and stop spending it on the first category.
Everything in between. This is where most decisions live and where nothing currently happens. Too consequential to guess, not consequential enough to justify six weeks. In practice these ship on the loudest opinion in the room.
That middle band is the actual problem. It is not that teams do not do research. It is that a large majority of decisions were never realistic candidates for research in the first place, so they get made on instinct and nobody counts them as a research failure because no research was ever planned.
What to do about the middle
Two things move the needle, and neither is more evangelism.
Make small evidence normal. Five unmoderated tests, a support ticket review, a session recording pass, a query against past studies. Each of these is hours rather than weeks, and each beats a guess. The cultural work is making it acceptable to bring imperfect evidence to a decision instead of nothing, which sounds obvious and is genuinely hard in organizations where research has been positioned as rigorous or not worth doing.
Reduce the cost of asking. The reason the middle band is empty is that the marginal cost of one more question is high. Anything that lowers it, including a synthetic panel grounded in evidence you already hold, changes which decisions can afford evidence at all. This is the honest case for simulated research and it is a narrower claim than the marketing usually makes: not that it replaces talking to users, but that it puts something in front of the decisions that were getting nothing.
The goal is not a hundred percent research coverage. That is not achievable and would not be a good use of anyone's quarter. The goal is that no consequential decision gets made with zero evidence purely because the only available instrument took six weeks.
The uncomfortable part for research leaders
If you run research and you want more of it to happen, the highest-leverage move is probably not arguing for headcount. It is shortening the distance between a question and an answer.
A function that returns something useful in two days will get asked far more often than one that returns something excellent in five weeks, and over a year the first one will influence more decisions. That is an unsatisfying thing to be true, because rigor is the craft and speed feels like a compromise of it.
But an instrument nobody reaches for has an influence of zero, however rigorous it is. The research that gets done is worth more than the research that would have been better.