How much market research actually costs

Market research is priced by method, not by output, which is why quotes for what sounds like the same project arrive an order of magnitude apart. A survey of 400 buyers, a round of twelve executive interviews, and a desk-research market assessment are three different cost structures wearing one name.

Here is what each actually costs, what drives the number inside a quote, and how to tell whether you are being charged for research or for overhead.

How much does market research cost?

For most business-to-business projects, expect roughly $5,000 to $50,000 for a scoped study, with full-service quantitative work running $30,000 to $100,000 and above. Desk-based market assessments sit well below that band; anything requiring recruited human participants sits well inside it, because recruiting people is the expensive part.

Those are working ranges rather than quoted prices. Treat them as a way to sanity-check a proposal, not as a benchmark to negotiate against. An agency quoting outside a band may be mis-scoped, or may be the only one who understood what you asked for.

What actually drives the price?

Four variables explain most of the spread between two quotes for the same brief. Understanding them is what lets you cut cost without cutting rigour.

Incidence rate is the dominant one, and the one clients consistently under-estimate. It is the proportion of people screened who qualify for your study. Surveying general consumers is cheap. Surveying hospital procurement leads at facilities over 300 beds who changed vendors in the last 18 months means screening hundreds of people per qualified respondent, and the screening is billed. A one-line change to your target definition can double a quote.

Sample size matters less than people expect, and follows diminishing returns. Going from 100 to 400 respondents roughly halves your margin of error; going from 400 to 1,600 halves it again at four times the cost. Very few business decisions need the second halving.

Number of markets and languages. Each additional country is close to a separate study: translation, local recruiting, local norms, and a separate analysis pass. Multi-market work does not benefit from the economies of scale clients assume it does.

How much thinking is included. This is the line that separates a $12,000 quote from a $40,000 one for identical fieldwork. Data collection plus a tabulation deck is one product; data collection plus an analyst who reconciles the findings against the market and tells you what to do is another. Both are legitimate. Confusing them at the quoting stage is how clients end up with 90 slides of crosstabs and no decision.

How much does it cost to hire someone to do market research?

If you are buying people rather than a project, the market is roughly tiered:

The in-house line is the one most often got wrong. A researcher's salary is not the cost of the function: panel access, survey tooling, and syndicated data subscriptions can add a substantial recurring line on top. That is the same arithmetic mistake as comparing a build estimate to a first-year licence in a build versus buy analysis.

Can ChatGPT do market research?

It can do the secondary research competently and cannot do the primary research at all. That distinction is the whole answer, and it matters more than any capability argument.

A capable language model with web access will find, read and synthesise published sources far faster than a human analyst: market structure, competitor positioning, public financials, regulatory context, published forecasts. This is genuinely the bulk of the hours in a desk-research engagement, and the cost implication is real.

What it cannot do is generate evidence that does not already exist. It cannot interview your customers, run a survey against a recruited sample, or tell you what buyers in your category believe this quarter. Ask a model to estimate a market it has no data for and it will produce a number, confidently, with no signal about how much to trust it. That failure mode is the reason for the provenance discipline in analyst-grade reports, where every figure is labelled reported, derived, or estimated.

The practical consequence for a budget is that the desk-research line should now be a great deal cheaper than it was, and the fieldwork line should not have moved at all. If a proposal prices both as though nothing has changed, ask why.

How do you get market research for free?

More is available at no cost than most teams use, and the limit is not availability but effort.

• Government statistical agencies. Census, labour, trade and industry data, generally the highest-quality free data that exists.
• Public company filings. Annual reports and investor decks contain segment sizing and competitive framing that vendors pay to reproduce.
• Trade associations. Industry surveys, often free to members and cheap to non-members.
• Analyst firm summaries and vendor reprints. Headline findings are published free; full reports reach you through vendors, with the selection bias that implies.
• Your own data. Sales objections, churn reasons, support tickets and won/lost notes. The most under-exploited research asset in almost every company, and the only one nobody else has.
• Your customers, directly. Twenty honest conversations cost nothing but calendar time and routinely outperform a commissioned study that nobody reads.

What should you actually pay for?

Cost discipline in research comes from scoping the question, not from squeezing the vendor. Three rules do most of the work.

Pay for evidence you cannot otherwise obtain. Primary fieldwork against a hard-to-reach audience is expensive because it is genuinely hard. Desk research is not, and paying agency rates for a literature review is the most common overspend in the category.

Size the study to the decision. Before commissioning anything, write down what you will do differently depending on the result. If no plausible finding changes the decision, the correct research budget is zero. If the decision is worth $2m, spending $40,000 to de-risk it is obviously sound, and spending $200,000 usually is not, because the extra precision does not change the choice. The same logic governs market sizing: a range that is decision-useful beats a point estimate that is falsely precise.

Buy the analysis separately if the fieldwork is commodity. Panel-based data collection is close to a utility. Interpretation is not. Teams that unbundle the two often get better analysis for less money, because they stop paying a full-service margin on the commodity half.

The uncomfortable summary is that most market research budgets are set before the question is written, which guarantees that the number is either wasteful or irrelevant. Write the question first. The price follows from it, and so does the answer to whether you needed to spend anything at all.