Why user research costs so much — and what lean looks like

Ask a team why they skipped talking to users before launch and you often hear: research is too expensive. They are not wrong about the invoices. They are often wrong about what they are paying for — and about whether the expensive version is the only version that works.

Where the money actually goes

Recruiting the right people

Finding participants who match your segment — role, experience, device, geography — is slow. Panels, screeners, scheduling, incentives, and no-shows are a business of their own. Agencies price that sourcing risk into every project. If your brief needs “enterprise admins who have used competitor X,” you are not buying clicks; you are buying a scarce calendar.

Facilitation is skilled labor

A good moderator designs tasks, runs the session, probes without leading, handles consent, and keeps the hour useful. That is expertise billed by the hour (or by the day). Watching someone use your product looks simple on a recording; keeping the session valid is the hard part.

Live time does not parallelize cleanly

One moderated hour is roughly one person’s attention. Eight interviews are not “one study” in labor terms — they are eight live blocks plus setup and teardown, often with a second person taking notes. Cost scales with sessions unless you change the format.

Synthesis is the iceberg

Recording is cheap. Turning hours of video into themes, clips, recommendations, and a deck stakeholders will trust is where days disappear. Buyers often underestimate this line item; vendors do not.

Overhead and theater

Project management, labs or travel, tooling seats, legal review, and polished reporting get packaged into “a research engagement.” Procurement also rewards artifacts that look serious — large n, long timelines, slide theater — even when a smaller iterative pass would have answered the decision sooner.

Why teams still pay it

Traditional research earns its keep when the question is open: new market, unfamiliar audience, early concept, political stakeholders who need a formal study. Breadth and facilitation quality matter there. The problem is using that machine for a narrower job — will someone get through this path before we ship? — and then concluding that all human evidence must cost five figures.

What lean looks like

Lean is not “skip humans.” It is cut everything that does not serve a named decision:

• One risk, one brief. Numbered success criteria instead of “explore freely.”
• A short form that mirrors the brief. Comparable answers beat essay contests.
• Small n on purpose. A few careful sessions, then fix, then another round — not one monster study.
• Evidence over narrative. Recording + marked moments + structured fields you can approve or reject.
• Iteration over theater. Budget for three sharp passes beats one expensive performance.

That pattern matches how productive teams already ship: tight QA, clear missions, artifacts that survive disagreement. It is closer to structured human testing than to a full generative research program — and it should be priced like focused labor, not like a quarterly research retainer.

What you should not cheap out on

• Independence — someone outside the build team still has to attempt the path.
• A real trail — without recording or marked proof, you are back to vibes.
• Honesty about scope — lean tests answer ship questions; they do not replace deep discovery when you truly do not know the problem space.

On QATested

QATested is aimed at that lean middle: projects that hold a product’s history, focused tests with short custom forms, paid testers, recordings, and approve/reject — human confidence without buying a full research agency package for every release risk.

Related: brief, high-signal tests · confidence from human testing · what human testing still reveals.