State of Insights 2026: Market Research Industry Trends From 126 Insights Leaders on AI, Data Quality, and Proving Impact
By Erica Dinger · Published 19 October 2026 · 16 min read
The biggest market research industry trend of 2026 isn’t AI. Among 126 senior insights leaders surveyed for the State of Insights: Insiders Edition, proving the business impact of insights ranked as the top priority for the next 12 months, ahead of every AI-specific concern. They have adopted AI broadly as an assistant, but 51% would not trust it to run any research task without human review.
Proving the business impact of insights is the top priority for the next 12 months among the 126 senior practitioners surveyed, ranked ahead of every AI-specific concern.
AI assistance is close to universal in research workflows (only 1% of respondents use it for none of the listed tasks), yet 51% would not trust AI to run any research task without human review.
Fraudulent respondents and AI-generated survey answers account for roughly 60% of top data quality concerns, and 28% of respondents discarded or re-fielded a study over data quality in the past year.
Insights teams are cited in leadership decisions more often than they measure their own impact: more than half report C-suite or board citations in the past quarter, but only 17% measure research ROI consistently.
The clearest next step for insights leaders is to make their judgment visible by documenting how evidence was verified, how stated intent was checked against behavior, and which decisions the work changed.
Insights leaders are heading into planning season facing a question leadership asks more pointedly every year: what did the research actually change? Budgets are tighter, AI tools are in nearly every workflow, and a recommendation built on fraudulent respondents or unmeasured influence is harder to defend than it used to be.
The State of Insights: Insiders Edition, published by Voxpopme with research partner MRII on September 30, 2026, shows where the profession stands. Its 126 respondents, most of them senior and long-tenured, rank proving business impact and data quality as their top two priorities, ahead of AI. They have adopted AI broadly as an assistant but draw a firm line at autonomy.
Voxpopme is the consumer insights platform that delivers customer intelligence and supports the expert judgment of insights leaders. This guide unpacks what the findings mean for teams whose work needs to hold up where decisions are made.
What Is the State of Insights: Insiders Edition?
The State of Insights: Insiders Edition is a study of 126 research, analytics, and insights practitioners, published by Voxpopme in partnership with MRII (Market Research Institute International) on September 30, 2026. It is the first Insiders Edition of Voxpopme’s State of Insights series, which Andy Barraclough introduced in Quirk’s earlier this year as entering its third year. This edition puts the questions directly to the people running research functions.
How the Study Was Run
Findings come from a combined base of 126 respondents across three sources: a core text-based survey of 100 respondents (79 full completes and 21 partial completes) and two supplementary video-response surveys of 16 and 10 respondents. Every included response passed each survey’s data-quality and fraud-screening checks. All multiple-choice and ranking findings use the full base of 126, and video answers add context to the numbers.
As we noted when we opened the study in August, we paired quantitative questions with a short video answer for each topic, because a percentage shows where respondents landed and a spoken answer shows how they got there.
Who Took Part
The sample skews senior and experienced. 56% hold Director-level roles or above (30% Director or Head of, 26% C-suite or VP), and 63% lead or manage the research function at their organization. 45% have more than 20 years in market research, insights, or analytics, while only 14% have five years or less. 93% work directly in market research or consumer and customer insights.
Respondents come from both sides of the industry. Brand or client-side organizations with a research focus and full-service research suppliers each account for 33%, followed by strategic consulting suppliers (27%) and qualitative research suppliers (21%). Many selected more than one category. The sample is 64% US-based, with 15% in the UK and the rest spread across Canada, Australia, and roughly 16 other countries.
How to Read the Numbers
These results reflect informed practitioner opinion from a senior, US-weighted group of 126. They describe what experienced insights leaders report, not a census of the profession. Multi-select questions don’t sum to 100%, so treat those figures as the share of respondents choosing each option rather than slices of a whole.
What Matters Most to Insights Leaders Over the Next 12 Months?
Proving the business impact of insights is the top priority, followed by data quality and sample integrity, then the evolving role and skills of the insights team. Respondents ranked their top three from a list of seven, weighted 3 points for first, 2 for second, and 1 for third.
| Priority | Weighted Score | Total Mentions |
|---|---|---|
| Proving the business impact of insights | 154 | 67 |
| Data quality and sample integrity | 133 | 63 |
| The evolving role and skills of the insights team | 119 | 64 |
| Research automation and orchestration | 113 | 57 |
| AI governance, ethics, and trust | 105 | 53 |
| Turning qualitative depth into quantifiable evidence | 62 | 34 |
| The say-do gap | 36 | 22 |
The two technology-led priorities sit fourth and fifth. For a profession whose conference agendas lean heavily on AI, that order says a lot about where practitioners feel the pressure. One detail is worth noticing: the evolving role of the insights team drew more total mentions (64) than data quality (63), even though fewer respondents ranked it first. Many leaders see their own role as a live concern, even if it isn’t the most urgent.
Andy Barraclough, Voxpopme’s CEO, frames it this way in the report’s closing analysis: “This is a profession thinking about its standing as much as its tooling.” The top three priorities share a root. Can the team trust what goes into the work, are people still the ones adding the value, and can anyone show that what came out of it changed a decision?
What Respondents Think Is Overhyped
42% of respondents said none of the industry’s current topics are overhyped. Among those who named one, research automation and orchestration came first (32 mentions, weighted score 81), followed by AI governance, ethics, and trust (60), the say-do gap (57), and turning qualitative depth into quantifiable evidence (57). Data quality drew just 8 mentions, the least likely of any topic to be called overhyped.
We read that as fatigue with the AI conversation, not with AI itself. Researchers are using these tools daily while building a new skill alongside them: checking what the tools tell them, since AI-generated output can look convincing even when it’s wrong. One anonymous respondent named the tension underneath it all: “Contradiction between increasing demand for consumer closeness and diminishing research budgets.”
How Are Insights Teams Using AI in 2026?
Almost every team surveyed uses AI to assist with at least one research task, but most keep a human reviewer in the loop, and very few would let AI run any task unsupervised. Only 1% of respondents use AI for none of the tasks listed.
| Research Task | Respondents Using AI Assistance |
|---|---|
| Text and open-end analysis (coding, tagging, thematic) | 68% |
| Synthesizing insights across studies | 54% |
| Report writing and charting | 49% |
| Survey data analysis | 45% |
| Survey design | 40% |
| Translating findings into business recommendations | 38% |
| Audio and video analysis, interview moderation | 28% |
| Prioritizing research questions | 16% |
| Sample sourcing and screening | 11% |
Where Trust Stops
Adoption is broad. Autonomy is not. 51% of respondents would not trust AI to run any research task fully without human review, and only 11% would trust it with any task unsupervised. The most-trusted unsupervised task is text and open-ended analysis, at 17%.
Add a human reviewer and the picture changes sharply.
| Research Task | Use AI Assistance Today | Would Trust AI Unsupervised | Would Trust AI With Human Review |
|---|---|---|---|
| Text and open-ended analysis | 68% | 17% | 58% |
| Synthesizing insights across studies | 54% | Not reported | 48% |
| Survey design | 40% | Not reported | 44% |
Survey design stands out. Slightly more respondents would trust AI there under human review (44%) than currently use it (40%), which suggests room for careful growth rather than resistance.
The pattern is consistent with MRII’s own tracking research. In its April 2026 study of 498 insights professionals, only about 1 in 10 said AI is fully embedded in their day-to-day workflows. The questions and samples differ, but both studies describe AI as present across the work without anyone owning it.
Respondents explain why in their own words. Dave, 63, describes AI for open ends as “fabulous,” particularly for interviews already conducted, “so you kinda know what was said.” Marjolein, 47, is candid about the learning curve: “I was not fully confident in the analysis; I still watch back interviews to be fully confident with the findings I was sharing.”
Why Expert Judgment Is the Line Respondents Protect
AI can code open ends, cluster themes, and pull evidence from across past studies. Expert judgment decides whether a theme is real, what it means for the business, and whether it’s strong enough to act on. That is the line most of these respondents are drawing. Ed Keller, Executive Director of MRII, makes the same point in his foreword, writing that human judgment “and the skills that surround it” are “becoming more important, not less.”
It’s also how we think about our own research agent, Voxpopme Compass, which we launched in May. Compass surfaces evidence from a team’s existing research and helps build the deliverable. The researcher decides what that evidence means and whether it’s ready for leadership. Note that 38% of respondents already use AI to help translate findings into business recommendations, the step closest to judgment. That is exactly where review standards matter most.
Data Quality Is the Biggest Named Threat
Data quality is the threat respondents take most seriously. 28% discarded or re-fielded at least one study in the past 12 months because of data quality concerns, 57% did not, and 6% were unsure. Most of those affected did it once or twice, but a handful re-fielded more than five studies.
When asked for the single biggest data quality concern right now, respondents named:
- Fraudulent, fake, or low-quality respondents, including bots: 41%
- AI-generated or LLM-assisted survey responses: 18%
- Panel over-use and respondent fatigue: 10%
- Sample representativeness: 8%
- Inconsistent quality across panel providers: 5%
Together, fraud and AI-generated responses account for roughly 60% of top concerns, well ahead of traditional issues like representativeness or panel fatigue.
An Old Problem in a New Form
Research has long fought bias and fraud. In 1936, the Literary Digest mailed around 10 million straw-poll ballots, received about 2.4 million back, and still called the presidential election for the wrong candidate. Later analysis in Public Opinion Quarterly traced the error to both a skewed sample and biased responses. Click farms threatened online panels in the early 2000s. Today’s version is harder to spot: bots and LLM-written answers that read as legitimate at first glance and quietly push results in the wrong direction.
The consequence connects directly to the AI findings. Analysis tools work with whatever they’re given, so a study built on invalid respondents produces confident-looking output that means very little. A discarded study is budget spent twice and a delayed decision, which is why data quality ranks second on the priority list and almost nobody calls it overhyped.
Respondents also want suppliers who can stand behind their sample. Nicholas, 29, put the frustration plainly: “It’s really hard to tell, and especially fishy when a sample provider will do quant for you, but they won’t recruit qual for you.”
Rigor starts here. A conclusion that holds up under senior scrutiny begins with respondents a team can vouch for. Formats that let a researcher see and hear the person responding add another review point, though they complement fraud screening rather than replace it.
The Say-Do Gap Many Teams Can’t Yet See
21% of respondents saw a stated-intent finding contradicted by actual customer behavior in the past 12 months, and a further 33% said they didn’t know, which suggests many teams aren’t set up to check. Where respondents did see the gap, purchase behavior (11 mentions) and product usage (9 mentions) came up most often.
| How Teams Validate Stated Intent Against Behavior | Respondents Using It |
|---|---|
| Behavioral or transaction data | 40% |
| Sales or CRM data | 30% |
| Observational or ethnographic research | 29% |
| Not applicable to my work | 18% |
| Nothing formal | 16% |
| Verified-buyer sample | 14% |
| Video-based evidence | 11% |
| Third-party panel or receipt data | 8% |
16% use nothing formal to validate what people say against what they do. The say-do gap ranked last of the seven priorities, yet it bears directly on the first. A recommendation built on stated intent that behavior later contradicts is exactly the kind of finding that weakens an insights team’s credibility with leadership.
Why the Goal Is Understanding, Not Catching People Out
The point of checking behavior isn’t to catch respondents in a lie. Most aren’t intentionally misleading anyone. People say healthy eating is their top priority and keep candy in the pantry. They say they’d definitely buy a new product in a concept test and never show up at the store. The gap is real, and the interesting question is why it exists.
Behavioral and transaction data can show that a gap exists, while asking customers about the discrepancy helps researchers understand why. Seeing and probing those discrepancies in video surveys opens new ways to understand how consumers see themselves and the products they buy. Only 11% of respondents currently use video-based evidence for validation, so most teams have room to pair the “what” from behavioral data with the “why” from customers themselves.
Proving ROI: Influence Without Measurement
Insights teams are cited in senior decisions more often than they measure their own impact. Only 17% of respondents formally measure research ROI or business impact consistently, while 32% do so sometimes, 21% rarely, and 12% never.
Meanwhile, more than half report their work was cited in C-suite, board, or leadership decisions in the past quarter. 23% cite 3 to 5 references, 17% cite 1 to 2, and 14% cite more than 10. Only 2% reported zero.
That is the credibility gap at the center of the report. Influence exists, but the evidence for it largely goes unmeasured and uncommunicated. A function whose reputation rests on being believed rather than demonstrated is exposed the moment budgets tighten. In the report’s closing analysis, Andy Barraclough puts it simply: “Judgment that cannot be seen is difficult to defend, and harder still to fund.”
Matt, 52, describes the problem from inside it: “I don’t doubt that it has value. I think the bigger challenge for us as an industry and as a business is being able to quantify that value and potentially help our clients to demonstrate the impact that research had on those business decisions.”
What a Visible Decision Trail Looks Like
One anonymous respondent shared an insight that visibly moved their business: “Past client promotions had concentrated on differences between business models and our research showed customers did not care. Product capacity was key, and subsequently the promotion messages changed.” That sentence holds everything an ROI case needs. It names a prior assumption, the evidence that challenged it, and the decision that changed as a result.
Capturing that chain, study after study, is how influence becomes cumulative. When past studies stay searchable, a team can trace a decision back to the customer evidence behind it, months or years later. It’s the thinking behind tools like Voxpopme Insights for Microsoft Teams, which returns answers from a team’s existing research with links to the customer videos behind each point, so the evidence shows up where decisions are being discussed.
What Should Insights Leaders Do With These Findings?
The common thread is visibility. Respondents are protecting the human judgment in their work, and the next step is making that judgment, and its effect on decisions, easy for leadership to see. Four moves follow from the data.
1. Treat Verification as Part of the Deliverable
With fraud and AI-generated responses topping the data quality list, a short note on how respondents were screened and how AI-assisted analysis was checked belongs in the readout, not just the methods appendix. It shows stakeholders the work behind the finding and gives the finding more weight.
2. Check Stated Intent on the Decisions That Matter Most
Not every study needs a behavioral check. High-stakes ones do, especially concept tests and purchase-intent work. Pair stated intent with behavioral, transaction, or CRM data where it’s available, and ask customers directly when the two disagree.
3. Keep a Simple Decision Log
Only 17% measure ROI consistently, and a formal model isn’t required to start. Record the decision each study informed, what the team believed beforehand, and what changed. Over a year, that log becomes the ROI case, built from evidence rather than recollection.
4. Define AI’s Role by Task
The trust data offers a practical map. AI assistance on open ends and synthesis, with human review, has majority or near-majority support. Interpretation and recommendation remain with people. Setting that out explicitly for a team removes guesswork and protects the part of the work that carries its professional value.
Where Consumer Insights Platforms Fit
Different tools do different jobs. Survey platforms suit structured measurement, panel and sample providers supply respondents, analytics tools model behavior, and full-service agencies bring capacity and specialist expertise. Voxpopme operates in the consumer insights platform category, capturing, analyzing, storing, and sharing customer evidence so it can shape decisions. Our approach centers on expert judgment: the platform helps researchers reach relevant evidence and keep it traceable, and researchers decide what it means.
Frequently Asked Questions
What is the State of Insights report?
The State of Insights: Insiders Edition is a 2026 survey of 126 research, analytics, and insights practitioners, published by Voxpopme in partnership with MRII on September 30, 2026. It covers priorities for the next 12 months, AI adoption and trust, data quality, the say-do gap, and how teams prove the business impact of their work.
Who took part in the State of Insights survey?
Respondents are senior and experienced. 56% are Director-level or above, 63% lead or manage a research function, and 45% have more than 20 years in the field. The sample includes both brand-side teams and research suppliers and is 64% US-based, with respondents from the UK, Canada, Australia, and other countries.
How are market researchers using AI in 2026?
Most respondents use AI as an assistant rather than an autonomous researcher. 68% use it to help with text and open-end analysis, and 54% use it to synthesize insights across studies. Trust depends on oversight: 51% wouldn’t let AI run any task unsupervised, while 58% would trust it with open-end analysis when a human reviews the output.
What is the biggest data quality concern in market research right now?
Among respondents, fraudulent or fake respondents, including bots, is the top concern at 41%, followed by AI-generated or LLM-assisted survey responses at 18%. Together they make up roughly 60% of top concerns. 28% of respondents discarded or re-fielded a study over data quality in the past 12 months.
What is the say-do gap in market research?
The say-do gap is the difference between what consumers say they’ll do and what they actually do. 21% of respondents saw a stated-intent finding contradicted by behavior in the past year, and 33% didn’t know. The most common validation method is behavioral or transaction data (40%), while 16% use nothing formal.
How do insights teams measure research ROI?
Inconsistently, according to respondents. Only 17% formally measure research ROI or business impact consistently, while 12% never do. Yet more than half say their work was cited in leadership decisions in the past quarter. A practical starting point is a decision log recording what each study informed and what changed.
Conclusion
AI has changed how research gets done, and the insights leaders surveyed have embraced it as an assistant. It hasn’t changed what research is for. The teams that earn and keep their seat at the table will be the ones that pair AI-assisted analysis with visible expert judgment, verified respondents, and a clear record of the decisions their evidence changed.
Get the full picture. Download the State of Insights: Insiders Edition to see every chart, the full respondent profile, and the practitioner voices behind the numbers.
Research partner: Ed Keller and MRII. Report collation and editing: Erica Dinger, Tom Higgins, and Finidi Lawson.
Share this post