How to Choose a User Research Agency: Key Questions

Before you shortlist: define the decision the research must inform

This guide assumes you have already decided to outsource. If you are still weighing that call, start with when to outsource vs build an in-house team and come back once it is settled.

Before you contact anyone, write down three things: the product or business decision the research must inform, the users you need to hear from, and what a useful result looks like. One page is enough. A tight brief lets you compare vendors on identical terms, and it flushes out the ones who take a vague ask at face value instead of pushing back. An agency that asks “what will you do differently depending on what we find?” is already doing you a favour.

Expect your shortlist to come together quickly. Buyers now build candidate lists through AI tools and peer referrals as much as through search, which is efficient but tends to surface the same well-marketed names. Arrive with your own evaluation rubric rather than a borrowed top-ten list. Our notes on how B2B buyers shortlist vendors today cover where that process goes wrong.

Questions about methodology and method fit

Ask which methods the agency has run in the last six months. Then ask the harder question: how do they choose between them for a given research question? The first answer tells you what they are comfortable with. The second tells you whether they think.

A good partner matches method to decision. Usability testing when the problem is in the interface and you need to watch where people stumble. Interviews when you need motivation, context, and the workarounds people have built. Surveys when you already know the questions and need scale to size a segment or an opportunity. Our comparison of usability testing vs user interviews covers where each earns its place. If every answer bends back to the same signature deliverable, you are being sold a product, not a study.

Probe how they use newer tooling. AI-moderated interviews can widen reach and cut cost for structured, exploratory conversations. A skilled human moderator still matters when the topic is sensitive, the follow-up probes are unpredictable, or the participant is a senior specialist who will disengage from a script. Ask when they reach for each, and read when AI-moderated interviews make sense to judge the answer.

Red flag: an agency that meets every question with the same method, or cannot explain plainly when qualitative depth beats quantitative breadth. For the wider landscape, see the full range of UX research methods.

Questions about participant recruitment and sample quality

Recruitment quality decides whether a study is worth running. Strong analysis cannot rescue the wrong participants. Ask who actually sources people: an in-house recruitment team, a specialist partner, or a marketplace panel. Ask how they screen. And get the no-show and replacement policy in writing — what happens when two of your eight sessions fall through on the day.

Request the screener itself. You are looking for criteria that verify identity, genuine category experience, and, for professional audiences, real credentials and job responsibilities. Hard-to-reach B2B roles are where weak recruitment shows. It is easy to find someone who claims to be a procurement lead for a mid-market manufacturer, and hard to confirm it. Ask what verification steps happen before a participant is confirmed.

Ask directly how they guard against fraudulent and AI-generated participants. Professional survey-takers, duplicate accounts, and scripted responses can degrade a dataset without any single answer looking wrong. Our guide to detect fake and AI-generated participants lists the checks a competent partner should already run.

Confirm panel provenance. Is the panel proprietary, third-party, or built from a list you supply? And who owns the participant relationships afterwards — can you re-contact willing participants for follow-up work, or does that access disappear when the engagement closes?

Questions about data ownership, consent and compliance

Get the ownership position in writing before you sign. You should own 100% of the raw data, the recordings, the transcripts, the analysis, and the final intellectual property outright — not hold a licence to material the agency continues to control. This sounds obvious until an engagement ends and you find the recordings sit behind someone else’s login.

Ask how consent is handled: how participants are told what the session is for, how personally identifiable information is redacted from transcripts and clips, how long data is retained, and how deletion is carried out and confirmed. Check that the process meets GDPR, or the equivalent in your region, and that it covers any sub-processors.

If AI notetakers or transcription services are used, ask where that audio and text are processed and stored, whether it feeds the training of third-party models, and how participants are told about it and consent to it specifically. AI notetaker consent and privacy covers the questions to put to a vendor here.

Red flags: deliverables that only open in the agency’s own platform, and vague answers about what happens to your data once the invoice is paid. Both are signs you are renting access to your own research.

Questions about deliverables and turning insight into action

Ask to see a real deliverable from a project like yours — a readout deck, a repository export, an opportunity backlog — not a polished case study written for marketing. You are judging clarity, structure, and how well the evidence supports each claim.

Look at how findings are framed. Prioritised design or product implications, tied to the decision you named at the start, are useful. A long, flat list of observations is homework you still have to do. Ask how they decide what rises to the top.

Ask what they do when results contradict what you believed going in. A partner worth keeping will show you the evidence and hold the line, not soften the finding to keep the room comfortable. Ask how they help you socialise results after handoff: a stakeholder session, a highlight reel, annotated clips people will actually watch. Our piece on turning research insight into product impact covers what good looks like.

Agree the deliverable format up front — readout, repository entries, highlight reel, opportunity backlog — so that “done” is unambiguous and you are not renegotiating scope during the readout.

Questions about pricing, scope and contracts

Ask for an itemised quote. Moderation, recruitment, incentives, analysis, and reporting are often priced as separate lines, and the gaps between them are where surprises live. Know exactly what is included and what is not — incentive costs and translation are common exclusions.

Benchmark the total against typical market rates so you can read the number in both directions. A quote well below the field usually means recruitment or analysis is being quietly cut. A quote well above it needs a reason you can see. Our guide to what user research typically costs in 2026 gives current ranges by method.

Clarify scope boundaries in the contract: how change requests are handled and priced, how many revision rounds are included, and whether payment is tied to milestones — a scoping sign-off, fieldwork completion, final readout — so incentives stay aligned through the project.

Read the IP and confidentiality clauses yourself. Do not assume they match the conversation you had. This is the one part of the contract where a five-minute read can save a serious argument later.

Red flags — and how a paid pilot de-risks the decision

Some warning signs are reliable. An agency that skips discovery and goes straight to a proposal. One that agrees to everything on the first call, including your timeline and your assumptions. A portfolio full of activity — sessions run, reports delivered — with no mention of what changed as a result. Evasive or shifting answers on data ownership. Any one of these is a reason to slow down.

The highest-signal check is a small, paid pilot: three to five interviews, or one round of usability sessions, scoped as standalone work. It shows you recruitment quality, communication cadence, and deliverable craft before you commit a full budget. Define what it must demonstrate: participants who match the brief, insight framed for a decision rather than a transcript dump, and honest updates when something slips.

Here is where that pays off. In a Jobs-to-be-Done study — work that maps the progress a customer is trying to make — for an early-stage online-learning company, we ran fieldwork as three iterative rounds rather than one large batch and treated the first round as a pilot. It caught a screener weakness early: a few participants had overstated how often they hit the problem we were exploring, and because the segments were built around frequency of need, that would have skewed the result. We tightened the screener and re-ran before the main rounds. A vendor who resists that kind of correction is telling you something. Our notes on research operations practices to expect from a partner describe the discipline that makes pilots work.

Always take two references from teams close to you in stage and domain. A strong enterprise partner can be the wrong fit for a seed-stage product.

Your user research agency selection scorecard

Turn the guide into a weighted scorecard with six criteria: method fit, recruitment quality, data and compliance, deliverable clarity, pricing transparency, and pilot performance. Weight them for your situation — recruitment quality matters more for a niche B2B audience; deliverable clarity matters more when findings must travel across a large organisation.

Score every shortlisted vendor on the same scale, from the same evidence, so the decision rests on what you saw rather than on brand recognition or who presented best.

Keep non-negotiables separate. If a vendor will not grant full data ownership, or cannot explain their fraud checks, that is an automatic disqualification regardless of total score.

Bring these questions to every vendor call:

• Which methods have you run in the last six months, and how do you choose between them?
• Who sources and verifies participants, and how do you screen out fraudulent or AI-generated ones?
• Do we own 100% of the raw data and IP, in writing?
• Can we see a real deliverable from a comparable project?
• What is itemised in the quote, and what is excluded?
• Will you run a small paid pilot first?

Your next step is simple: build the scorecard before the first call, not after the last one.

Frequently asked questions

How much should I budget to hire a user research agency?

It depends on method, how hard the audience is to reach, and scope. A round of moderated usability testing with mainstream consumers sits at the lower end. In-depth interviews with senior B2B specialists, or multi-market studies, cost considerably more because recruitment and incentives climb. Get an itemised quote and benchmark it against the ranges in what user research typically costs in 2026 rather than working from a single figure.

What questions should I ask a user research agency before hiring?

Work through six areas: method fit (how they choose methods, not just which they know), recruitment quality (sourcing, screening, fraud checks, no-show policy), data ownership and compliance (you should own everything outright), deliverables (see a real one from a comparable project), pricing and contract (itemised quote, IP clauses), and references (two teams like yours). Then ask whether they will run a small paid pilot before the full engagement — the answer, and the pilot itself, tells you more than any pitch.

---

About Glasgow Research — Glasgow Research helps B2B SaaS teams turn customer and market research into product decisions. Work with us.