Market Research Statistics That Matter in 2026

Last updated: 22 September 2026

Quick Answer: The market research industry sits inside a global insights industry that turned over $153 billion in 2024, up 8%. ESOMAR's 2025 edition is the current audited count, and no 2026 edition has been published yet. Most other figures quoted in roundups are older, narrower, or both.

Every figure below was checked against its primary source in September 2026, and each one carries the edition it came from. Where a current edition could not be found, the number was removed along with the sentence that carried it.

The load-bearing correction is the headline number itself. The widely repeated $140 billion for 2024 and $150 billion for 2025 do not match ESOMAR's own series, which puts the insights industry at $153 billion in 2024 after an 8% rise from about $142 billion the year before.

That is why this page carries fewer numbers than the roundups it replaces. The ones that survived can be quoted without a caveat, which is the only property a statistic needs.

The table runs from industry-wide figures, through the population a sample is drawn from, to the method and data quality figures, so the order carries no ranking. The Year column is the year the figure was measured, not the year it was published.

How Big Is the Market Research Industry in 2026?

The global insights industry turned over $153 billion in 2024, an 8% rise on the year before, according to ESOMAR's Global Market Research 2025 report. That is the most recent audited total, built from national association data across more than 110 countries and regions covering about 80% of the global industry. A 2026 edition of the same report had not appeared on ESOMAR's publications index as of 22 September 2026.

The figures worth keeping:

• $153 billion in 2024 turnover, covering market research, research software and business services together.
• 8% growth, from about $142 billion in 2023, measured rather than projected.
• 952,700 people employed in the United States as market research analysts and marketing specialists in 2025.
• 7% projected growth in that occupation between 2025 and 2035, adding 66,300 posts.
• 82,000 openings projected each year over that decade, most of them replacing people who leave.

Two cautions. The occupational count is a wider net than the agency headcount roundups usually quote, because it includes marketing specialists. And ESOMAR revises its series annually, so a number lifted from a 2023 edition will not reconcile with the current one.

Which Companies Lead Market Research Revenue?

ESOMAR's Global Top Insights ranking, published free in its 2026 edition, is the reference ranking of the largest players across market research, research software and business services. Per-company revenue sits inside the report itself, and the per-company figures circulating in statistics roundups do not survive a check against a primary filing.

The clearest example is Salesforce, listed in several roundups at $4.52 billion of research revenue. Salesforce's own 10-K reports $5.99 billion of research and development expense for the fiscal year ended 31 January 2026. That is a software engineering budget, not money earned running studies, and the two are not interchangeable.

The Gartner, IQVIA, Nielsen and Adobe entries in the same lists date from 2021 and 2022 and have not been refreshed since. They are left out here rather than restated with a stale year attached.

For buyers the ranking answers a different question than they are usually asking. If the task is shortlisting rather than counting, the comparison of market research software platforms is the more useful page.

Where Does Research Spending Actually Go?

Spending divides three ways, and the division matters more than any single share. ESOMAR splits the insights industry into market research, research software and business services, and its forecasting work finds research software the primary growth engine while the market research sector grows more slowly and navigates data quality pressure.

The per-method revenue shares that circulate widely, 35% to online and mobile quantitative, 20.2% to reporting, 6% to online qualitative, come from paywalled editions and could not be checked against a current one. They are not repeated here.

What the public record does support is the direction of travel. Software and platform spending is growing faster than spend on studies, which squeezes the agency middle and pushes buyers toward full-service managed research at one end and self-serve tooling at the other.

That split is also why method choice has become a budgeting decision rather than a methodological one. The types of market research each sit in a different part of that cost curve.

What Do Measured Response Rates Show by Survey Mode?

Mixed mode wins on response rate, reaching 39.3% against 32.5% for mobile web alone and 22.4% for telephone. Self-reported usage surveys are the weakest evidence in this field, because researchers report what they intend to run. A randomized trial published in BMJ Open in March 2026 measured what actually happens: 4,800 patients discharged from four general hospitals in South Korea, surveyed between October and November 2022, randomly assigned to telephone, mobile web or a mixed mode, five contact attempts each. The fieldwork year is the one that matters when the figures get quoted.

Ordered by response rate, lowest first, because that is the variable the trial set out to move.

Contact rates were comparable between telephone and mobile web. The gap was cooperation: people who were reached agreed far more often on mobile. That is the mechanism behind why telephone response rates collapsed, and it also explains why mixed mode beat both single modes.

The practical reading is that mode is a recruitment decision before it is a data decision, which is the step most market research plans skip.

How Much Research Is Now Mobile First?

Mobile is the default channel rather than a channel option. GSMA Intelligence counted 5.78 billion unique mobile users in October 2025, 70.1% of the world's population, and Ericsson data in the same Digital 2026 report put smartphones at 86.9% of handsets in use.

• 5.78 billion unique mobile users, 70.1% of the global population, October 2025.
• 86.9% of handsets in use are smartphones, which leaves roughly one in eight that is not.
• 91% of US adults own a smartphone and 98% own a cellphone of some kind.
• 96% of US adults use the internet and 78% have home broadband.

The residual matters more than the headline. Alchemic runs AI-moderated interviews natively inside WhatsApp, with no link and no app, and outbound AI phone interviews to any working number, feature phones included, which reaches the part of a sample that a browser link leaves out. The mechanics of the first route are set out in how WhatsApp qualitative research works.

What Do the AI in Market Research Numbers Show?

There is no current, independently audited count of AI adoption inside the research industry. The percentages in circulation, 83% planning to invest and 47% using AI regularly, come from industry self-report surveys whose current editions are paywalled, so they are not quoted here.

What is on the public record is adoption in the wider population and governance inside the profession:

• More than 1 billion people use large language model and generative AI platforms each month, on Kepios analysis of October 2025 data.
• ESOMAR launched its AI Alliance on 14 July 2026, expanding its earlier AI Taskforce into three workstreams covering ethics and code, best practice, and training.
• ESOMAR's 2026 checklist, 10 Questions Buyers Should Ask Suppliers of AI-Based Research Services, replaced the buyer-facing half of the older 20 Questions and added an optional green, amber and red framework for grading supplier answers.

Governance moved faster than measurement this year. For what the tooling actually does today, see AI market research tools for consumer insights.

How Reliable Is Synthetic Data in Research?

Reliability depends on what is being measured, and published accuracy runs from near-perfect to near-chance. A September 2026 evaluation paper argues that the spread reflects differences in what is measured rather than differences in method quality, and that aggregate measures can look strong while hiding a complete absence of respondent-level differentiation.

The measured results:

• Across 108 attitude questions from a nationally representative survey of 3,063 people, screening digital twins at an R-squared above 0.7 raised the mean twin-to-human individual-level correlation by 15%.
• The same screen cut the share of poorly answered questions from 25.9% to 4.3%.
• A separate 2025 arXiv preprint on semantic similarity rating reached 90% of human test-retest reliability across 57 personal care surveys and 9,300 human responses.

For individual-level inference a fielded human sample is still the better choice, and the screening step above exists precisely because most questions fail it. Vendor claims in this category are worth reading against synthetic respondent platforms before any of it reaches a decision.

Where Does Research Budget Actually Leak?

Budget leaks at the cleaning stage, before the data reaches analysis. The client-sector splits quoted everywhere, pharmaceuticals at 16.6% and media at 15.5%, sit inside ESOMAR's paywalled report and no current edition of those shares could be read, so the sector answer is unavailable. What is measurable, and more useful to a budget holder, is the removal rate.

The Insights Association's Global Data Quality benchmarking, H1 2026 wave, drew on 13 countries and nearly 1.8 million survey records:

• 28.5% combined pre-survey and in-survey respondent removal rate reported by research agencies.
• 21.2% for the same measure reported by sample suppliers.
• B2B studies carried the highest combined removal rates and exceptionally high post-survey cleanout.
• Link encryption adoption reported between 46% and 50%, with nearly half of all records not reporting the metric at all.

Roughly a quarter of the records entering a study never reach the dataset, though the pre-survey blocks inside that figure are screened out before a buyer pays for them. That is the real sector story, and it is the one that makes the case for why market research is worth the spend harder to argue on volume alone. Alchemic delivers a 200-interview qualitative study in about three days, with managed fieldwork or bring your own recruitment across 14 markets including the USA and the UK.

What These Statistics Cannot Tell You

None of these numbers tells you whether one particular study will actually work. Industry size, mode response rates and adoption counts describe averages across countries, categories and populations, and a single brand's decision sits inside variance that the averages hide.

Mode effects and mode selection are separate problems. People answer the same question differently by mode, and different people show up by mode. Conditioning on mode reduces the first and may introduce collider bias where the second exists, a consequence an American Journal of Epidemiology paper said in May 2026 may be underappreciated in epidemiology.

Simulated respondents recover countries, not people. Tested against European Social Survey Round 11, 30 countries, 42 items, adding a country name to a three-variable backstory lifted the median per-item correlation between simulated and observed country means from -0.03 to 0.52. An average of neighboring countries, using no model, recovered country levels more accurately than every model condition.

Estimates get revised. The Facts and Figures 2025 release puts 6 billion people online in 2025, about three-quarters of the world, and in the same release ITU restated its 2024 internet-user figure from 5.5 billion to 5.8 billion, a 300 million swing in a number many decks had quoted. The direction of market research trends outlasts any edition's decimal place, and so does the market intelligence behind a decision.