Here is an uncomfortable accounting exercise: count the product decisions your team shipped last quarter, then count how many of them were informed by any research at all. For most teams the ratio is brutal, and it is not because anyone is lazy or incurious. It is because of price.
A traditional recruited study has a cost structure that has barely moved in twenty years: recruit a panel, schedule sessions, run them, synthesize, present. Weeks of calendar time, thousands to tens of thousands of dollars, and a meaningful slice of a researcher's quarter, per question. At that price, only the biggest questions clear the bar. Everything else, the empty-state copy, the onboarding step order, the pricing page framing, the error message, ships on instinct. Teams do not choose to skip research on ninety percent of their decisions. The price chooses for them.
What the price collapse changes
Simulated research changes the marginal cost of a directional answer. In our own comparisons against a traditional recruited study cycle, a grounded simulation pass runs on the order of fifty times cheaper and turns around roughly twenty-eight times faster for equivalent directional coverage. Both numbers describe a directional screening pass, not a like-for-like replacement for a recruited study; the point is not that the two instruments are equal, but that one of them now costs so little that the demand curve changes shape.
When something that cost weeks starts costing an hour, the interesting effects are never just "the old thing, faster." They are the new behaviors the old price made impossible:
- Decisions that never got evidence start getting it. Not because anyone mandates it, but because asking became cheaper than arguing. The threshold question shifts from "is this worth a study?" to "is this worth an hour?"
- Research moves from episodic to continuous. The unit of research stops being the study, commissioned when a question grows big enough, and becomes the standing model of your users that any question can be run against at any time. Studies are events; a world model is infrastructure.
- Evidence compounds instead of expiring. A study's findings decay in a drawer. A grounded panel gets more trustworthy with every shipped outcome that is fed back into it, because every prediction that meets reality is a calibration point. The asset appreciates with use.
- The sequencing of building changes. When directional evidence arrives in an hour, you can afford to check before you build rather than instrument after you ship. Watching a dashboard after launch is research on your real customers, performed after the decision already hardened. Simulation moves some of that learning to before the commit.
What happens to researchers
Every prior cost collapse in a professional field has produced the same anxious question, and the same eventual answer. When making things becomes cheap, the people who understood the expensive version do not disappear; their judgment gets applied at a higher level, across far more decisions than they could previously touch.
The researcher's leverage in this new cost structure concentrates in three places. Deciding what evidence grounds the panel, because grounding quality is the ceiling on everything downstream. Designing studies and scenarios that produce decisions rather than flattery. And validating, running the backtests and the recruited confirmations that tell the team where the instrument can be trusted and where it cannot. One researcher operating that loop can put evidence behind a volume of decisions that would previously have required a department. That is not the elimination of a role. It is the removal of its bottleneck.
The risk that comes with cheap answers
The honest cost of this shift is a new failure mode: evidence theater. When answers are expensive, teams at least know when they are guessing. When answers are cheap, a team can wallpaper every decision with simulated findings and mistake volume of output for quality of knowledge. A fluent, well-formatted, wrong answer is more dangerous than no answer, because it arrives wearing the costume of diligence.
The defenses are unglamorous and non-optional. Ground panels in real evidence and treat ungrounded output as speculation. Prefer rank orders and distributions over point estimates. Backtest against decisions whose outcomes you already know. And keep recruited research in the loop where stakes are high, per the hybrid workflow we use ourselves. Cheap directional evidence plus disciplined validation is a genuine upgrade to product judgment. Cheap directional evidence alone is just faster guessing.
The bar moves
The teams this shift rewards are not the ones that generate the most options; generation is becoming table stakes for everyone. They are the ones that can tell, before shipping, which options deserve to exist, and whose sense of their users compounds instead of resetting with every reorg and every stale slide deck.
Expensive research does not disappear in that world. It gets promoted. Freed from carrying every routine question, it concentrates where nothing else will do: the novel markets, the irreversible calls, the human conversations that seed the next generation of grounding. The six-week study stops being the tax on every decision and becomes what it always should have been, a precision instrument pointed at the questions that deserve it.
That reallocation, more than any speed multiple, is the real economic story. We wrote up the longer argument for where judgment goes when building becomes abundant in our thesis.