Costs
Part of Primary market research: the version that survives contact with reality
How to choose a primary market research agency
Choosing a primary research agency means scoring the sample, the consent path, the raw data, and the incentive plan before you sign a fielding contract.
What to take away
- Give every primary market research agency the same minimized brief, then score the named team, the sampling frame, the consent path, the instruments, and the raw data. Keep the exit terms on the same scorecard.
- Ask who recruits participants and from what frame. A panel that cannot describe its source cannot be checked for bias.
- Match provider type to the jobsample exchanges price per interview, platforms sell subscriptions, recruiters charge by the session.
- Budget a paid pilot on consequential work. It is the only cheap way to see the invitation, the screener, and the coded data before you commit.
- Keep the evidence file next to the decision so a reviewer can rebuild your reasoning without asking you.
What the brief has to settle first
Write the business question before the request for proposals. A study built to answer a launch decision needs different methods than one built to size a market.
Match method to decision
Need reactions to concepts?
Focus groups
In-depth interviews for sensitive audiences
Decide the method yourself and let agencies price it. Focus groups suit reactions to concepts. In-depth interviews suit sensitive or expert audiences. A quantitative survey suits measurement across a defined population. Agencies that propose all three without asking about your decision are selling capacity, not fit.
Then set the sampling frame. Say which population you need, which markets, which languages, and which groups you will exclude. A frame you cannot describe is a frame you cannot defend later.
Hanover Research
Hanover Research, based in Arlington, Virginia, describes qualitative work in focus groups, interviews, digital ethnography, and diaries. Its qualitative page also lists recruitment, facilitation, coding, and reporting.
Its corporate research page describes surveys, interviews, ethnographic work, modeling, and advisory support.
That combination suits a team that wants one relationship across several studies. Get the assigned researchers named in the contract, not the pitch. Ask which recruitment sources feed each market and what is subcontracted.
Ipsos
Ipsos, founded in 1975 and listed on Euronext Paris, publishes a solutions catalog covering qualitative, quantitative, panel, analytics, and social listening. It fields work in about 90 countries.
That reach suits multi-market briefs where one vendor fields the same instrument in several languages. Breadth is not fit. Confirm the contracting entity, the local team, which panels or partners are used, and how the instrument stays comparable across markets. Translation drift is the usual failure.
Kantar and YouGov: panel owners that run their own studies
Kantar, based in London, gives buyers panel access through Kantar Profiles and household purchase tracking through Worldpanel. Kantar Marketplace sells automated tests at set project prices.
YouGov, listed on London's AIM market, says its panel spans more than 50 markets. It quotes most work per completed interview, which suits repeated tracking studies where panel continuity matters.
Dynata, Cint and Toluna: sample you can buy by the interview
Dynata, based in Plano, Texas, sells first-party panel sample and full-service fieldwork. Cint, listed in Stockholm, runs a sample exchange that returns feasibility checks and per-interview pricing. Toluna runs a panel in dozens of markets plus Toluna Start, a self-serve tool that quotes automated study pricing.
Sample-only suppliers charge by the complete. General-population completes in the US typically land between $1 and $5. Narrow B2B or physician audiences typically run $25 to $150 or more, depending on incidence and screening.
Qualtrics, Zappi and SurveyMonkey Audience: platforms you run yourself
Qualtrics, based in Provo, Utah, sells annual platform licenses and runs fieldwork through its research services team. Zappi automates concept and creative testing for consumer brands, often returning results within a day, sold as a platform subscription.
SurveyMonkey Audience sells responses by the complete. General-population pricing typically starts near $1 to $3 per complete and rises with targeting.
These platforms suit teams that will write their own questionnaire. You own the sampling logic, the consent text, and the analysis.
Sago: qualitative recruitment and facility work
Sago, formerly Schlesinger Group, recruits qualitative participants and runs research facilities in US and European cities. It suits healthcare, B2B, and in-person sessions you watch from behind the glass.
NORC, Westat and RTI International: public-sector frames
NORC at the University of Chicago runs the General Social Survey and other federally funded studies. Westat, in Rockville, Maryland, is employee-owned. RTI International sits in Research Triangle Park, North Carolina.
These suit studies that need probability samples, federal reporting standards, or review board oversight. Expect longer timelines and a higher cost per complete than a commercial panel.
Score the sample and the consent path
| Dimension | Evidence to request | Red flag |
|---|---|---|
| Team | Named researchers, roles, comparable studies | Only the pitch team is named |
| Sample | Frame, source, exclusions, incidence | Representation claimed without a source |
| Consent | Screener, consent text, incentive terms | Consent described but not shown |
| Data | Raw exports, coding frame, deletion terms | Data returned only as slides |
| Price | Cost per complete, change orders | One lump sum, no unit breakdown |
| Exit | Ownership, publication rights, deletion | Vendor can reuse your data |
Recruitment is where quality is won or lost. Ask for the source of the frame, the incidence rate the vendor assumes, and how they handle a market that under-recruits. An agency that cannot answer will substitute convenience for representation.
Score the vendor before signing
- Teamnamed researchers, roles, comparable studies
- Sampleframe, source, exclusions, incidence
- Consentscreener, consent text, incentive terms
- Dataraw exports, coding frame, deletion terms
- Pricecost per complete, change orders
- Exitownership, publication rights, deletion
Consent is a document, not a promise. Ask to see the screener, the consent language, and how incentives are disclosed. For anything touching personal data in California, the CCPA and CPRA set the rules and your counsel interprets them. Name the regulation to the vendor and let your lawyer read the paperwork.
Pilot before you field
Run a paid, time-limited pilot on any study that will inform a real decision. Ten to twenty completes is usually enough to expose a broken screener or a leading question.
Read the invitation, the consent text, the questionnaire, and the first coded responses. Then ask what the agency would change. An agency that defends a flawed instrument in the pilot will defend it in the report.
For secondary figures, start with the U.S. Census Bureau and BLS data before you pay anyone to collect what already exists. Primary market research: the version that survives contact with reality should answer what published data cannot.
Check the claims you will repeat
The FTC advertising substantiation policy requires a reasonable basis for objective advertising claims before they run. If a study will support a public claim, the method has to survive that standard, and your counsel decides what that means for your facts.
The FTC guidance on soliciting and paying for reviews covers fake feedback, selective requests, conditioned incentives, and undisclosed relationships. The same discipline applies to research testimonials and case studies you publish.
Common questions
Is this a ranking?
No. It is an unpaid starting point built from official service descriptions from each firm, dated August 31, 2026. Providers appear in groups by business model, and none paid for placement.
Why these providers?
A shortlist grouped by business model is easier to evaluate deeply than a long alphabetical list. Add local or specialist firms against the same scorecard.
Should we pay for a pilot?
Yes when the decision is consequential. A bounded pilot tests recruitment, participant care, instrument quality, and how the agency handles corrections.
Who owns the data?
Ask before you sign. Get exports, coding frames, and deletion terms in the contract, and confirm what happens to participant records at the end. What primary market research asks, and why walks through the questions to put to any vendor before you commit.
