Market Research in Flux: Four Forces Reshaping How Insights Get Done
By Brian Lamar, Sr. Director of Sample Quality, ROI Rocket
The market research industry hit $165 billion in 2025*. That’s up $12 billion year over year — not bad for an industry that’s supposedly being disrupted out of existence. But the headline number obscures something more complicated underneath: this isn’t a growth story, it’s a restructuring story. The money is real, but so is the volatility. And the two are not unrelated.
Andrew DeCilles and I recently hosted a webinar where we tried to map what’s actually happening — drawing on conference content from Samplecon, IA Ignite, and IIEX, published industry benchmarks from GRIT and GRBN, and a lot of conversations with clients and partners who are navigating this in real time. What follows is the substance of that conversation.
The four forces — and why they’re a system, not a list
Every trends conversation risks turning into a list of disconnected observations. “AI is big. Fraud is a problem. Capital is flowing in.” All true. All unhelpful on their own.
What’s actually happening is that four forces are operating on each other in ways that create feedback loops. Understanding those loops is more useful than tracking individual trends, so let me walk through each force and then show you where they converge.
1. Capital influx: the accelerant
Private equity and venture capital have been circling market research for years. What’s different now is the pace and the source.
Q1 2026 alone saw $861 billion in global M&A activity. Within the insights space: Aaru hit a $1 billion valuation on synthetic AI respondents. Simile raised a $100 million Series A from Index Ventures. UserTesting acquired User Interviews. Discuss merged with Voxco. Rep Data acquired ReDem and SightX. That’s just a partial list from the past few months.
Two announcements from May deserve special attention. OpenAI launched what it’s calling “The Deployment Company” — a $10 billion entity backed by TPG, Bain, SoftBank, and others — with an explicit mandate to embed AI into the operational systems of PE portfolio companies. Anthropic did the same thing with a $1.5 billion joint venture backed by Blackstone, Goldman Sachs, and Hellman & Friedman, sending applied AI engineers directly into businesses to redesign workflows around Claude.
These are infrastructure plays. And the companies they’re targeting — mid-market businesses under PE ownership — are exactly where most of our clients’ insights teams sit.
The structural implication: capital changes incentives, not just ownership. When PE takes over a research firm, the 3-to-5 year exit horizon and the EBITDA target don’t coexist neutrally with what good research requires. Quality infrastructure doesn’t show immediate ROI. Saying no to a project that can’t be fielded properly is a margin drag. QA headcount looks like overhead. Every one of those things is under pressure from day one. The tell is always the same: what gets cut first after an acquisition? It’s never sales. It’s account management, quality assurance, and methodological depth.
For buyers, the question isn’t just “does this vendor do good work” — it’s “who owns them, what do their investors need from them, and what happens to my data quality when they hit their Q3 targets.” That sounds cynical. It’s just informed procurement.
2. The data quality crisis: the breaking point
The GDQ/Insights Association recently released benchmark data that the industry should be sitting with. These aren’t estimates — they’re anonymized data contributed by member companies.
For general B2B studies, the post-survey cleanout rate is 59.6%. Nearly six in ten completed surveys are discarded after cleaning. For general B2C, it’s 13.1%. For healthcare patients, 30.7%.
The B2B number in particular is not a vendor problem. It’s a structural credibility problem that the industry has measured, documented, and published — which means it’s also an open invitation to anyone who wants to argue that traditional survey research is broken.
The GRBN Online Buyers Sentiment Survey adds the human dimension. Among researchers surveyed:
- 63% accept fraud as unavoidable
- 64% have personally experienced fraud
- 56% say fraud has influenced a business decision
- 39% trust the industry
The 39% figure is the one that should be keeping people up at night. Not because researchers are untrustworthy — but because the inputs to their work are. When a problem is pervasive enough and the tools to fix it are either expensive, imperfect, or inconvenient, smart people rationalize it. They inflate sample sizes as a hedge. They start treating garbage data as the cost of doing business.
That rationalization is exactly the opening that synthetic data companies are walking through. Not necessarily because their product is proven — but because the alternative has a documented credibility problem.
The so-what here is uncomfortable but simple: you can’t fix trust with a disclaimer. You fix it with provenance. Chain of custody. Verified respondents. The ability to say, for every complete in this dataset, here’s who they were, how we confirmed it, and what we did when something looked wrong. That’s what turns sample quality from a line item into a differentiator.
3. AI disruption: the disruptor
At IIEX North America 2026, ROI Rocket catalogued 29 companies across seven AI technology categories — 14 of them AI interview platforms alone — with the top five having raised more than $89 million combined. A year ago, maybe half of firms in our network were seriously thinking about AI integration. Now everyone is having the conversation.
The categories worth paying attention to:
AI-moderated interviews — Voice and text AI conducting hundreds of simultaneous in-depths and synthesizing themes in minutes. Outset, Strella, Dialogue AI, and Riley AI are the names getting the most attention. The key question to ask any of these vendors: do they have a chief research officer or methodologist? Are they a tech company that discovered research, or a research company building better tech? The answer shapes how you should evaluate their output.
Synthetic respondents — AI models anchored to verified panel data to augment or replace fielding, primarily for niche B2B segments and hard-to-reach populations. The honest framing here: synthetic respondents are solving a reach problem, not a replacement problem. The use case that makes genuine sense is where incidence is so low that traditional fielding is prohibitively expensive or structurally impossible. The mistake is deploying synthetic where human sample is available and verifiable — and too many teams are doing that right now because it’s cheaper, not because it’s better.
Every credible synthetic platform anchors to a verified human baseline. Aaru, Qualtrics Edge, Simile — none of them claims the model works without quality human data. The narrative that synthetic replaces human is almost always a funding deck, not a methodology paper.
Agentic research assistants — AI agents that design, field, clean, analyze, and report end-to-end from a plain-language business question. This category doesn’t work perfectly today. But when it does work, it changes who can do research — which is either an opportunity or a threat depending on where you sit.
The LinkedIn B2B Institute published relevant research on this. Among GenAI power users, they found three counterintuitive patterns: people are 25% more likely to use AI for tasks they already enjoy (AI amplifies motivation, it doesn’t manufacture it); GenAI optimists are 15% more likely than pessimists to be worried about job displacement (informed concern is not the same as pessimism); and high performers are 2.3 times more likely than low performers to fear that AI is eroding their market differentiation.
4. Methodology reframe: the response
The qual/quant wall has largely dissolved. AI makes it economically viable to run hundreds of in-depth interviews and analyze open-ends at scale. Methods that used to be either-or are now both-and.
The firms that understand this are building hybrid capability. The firms that don’t are watching their margins compress as AI platforms offer faster and cheaper alternatives to pure-methodology work.
Qualitative researchers need to understand tech, advanced data, and proper sampling. Quantitative researchers need to understand how qual supplements quant and answers the “why” questions that numbers can’t reach on their own. Neither can afford to stay siloed.
The practical implication: build proprietary data assets now. Pure-methodology firms have no defensive moat. Pilot hybrid qual-plus-quant before your competitors do. The firms that move first will set the standard that everyone else has to meet.
Where the forces converge
These aren’t four separate trends. They’re a system.
AI disruption creates pressure on methodology, which creates pressure on qual and quant providers to converge. The quality crisis makes every traditional approach look fragile, which creates an opening for AI-native entrants, which attracts more capital, which compresses the timeline on everything else. Capital funds both the AI platforms and the consolidation of incumbents — which reduces independent quality oversight — which accelerates the quality crisis.
The methodology reframe is the output. It’s where all three of the other forces arrive. And the industry doesn’t get to choose whether it happens — only whether it leads or follows.
If you’re waiting for one of these forces to stabilize before making a strategic move, you’re waiting for something that won’t happen. The stable state is in the rearview mirror.
What the disruption narrative gets wrong
One thing worth pushing back on: the loudest voices at conferences are not the industry. The firms co-presenting with major brands and announcing nine-figure valuations represent one visible slice. The industry is still mostly the people who just do the work.
Synthetic data has a lot of excitement behind it. It has very little consistent, believable proof. Jay Scruggs had a LinkedIn post recently that put it well: the synthetic data conversation has gotten ahead of the evidence. Plenty of excitement. Almost zero believable proof. Most traditional researchers agree. Holding that line is part of our job as insights leaders.
Qual isn’t dying either — it’s becoming the differentiator. When every quant study runs faster and cheaper via AI, the thing that becomes scarcer and more valuable is genuine human depth. Researchers who can moderate, probe, and interpret — and connect that to the numbers — are harder to replace than the ones who just field surveys.
And the trust problem won’t be solved by more technology. 39% of researchers trust the industry. That’s not a tooling problem. It’s a culture problem. New platforms deployed faster into a low-trust environment produce lower-quality decisions at higher speed.
Some practical takeaways
For insights leaders specifically, the strategic picture is dividing. The industry is splitting into tech-led firms (growing, integrated, margin-protected) and service-led firms (margin-compressed, losing share). The middle is collapsing. There’s no neutral position.
A few things worth doing now:
Pick your lane. Become the verified-quality specialist, the AI-native disruptor, or the deeply embedded operational partner. Generalist mid-tier shops are getting squeezed from both directions.
Build a deployment story, not just a research story. Buyers increasingly want to know how insights become decisions become outcomes — not just what the data says.
Invest in proprietary data assets. Verified panels, behavioral feeds, identity-validated samples. These are the assets that create defensible distance from competitors.
Treat your AI strategy and your trust strategy as the same conversation. The firms that lead on AI without leading on data quality will get caught when their first major fraud incident surfaces publicly.
Make sample quality actionable. Ask partners about custody and provenance, not just methodology. If a vendor can’t tell you where their B2B completes come from and what happened to the 59.6% they threw out, that’s an answer.
The research industry is being rebuilt in real time. The capital is real, the fraud numbers are real, the new technology is real, and the methodology pressure is real. What’s also real: the researchers who understand what AI cannot fabricate — a verified person, with genuine opinions, giving honest feedback — are going to be the ones who matter most when the dust settles.
That foundation is more fragile right now than most people want to admit. Protecting it is worth treating as a strategic priority.
Brian Lamar is Sr. Director of Sample Quality at ROI Rocket, a MarketOnce company. He co-hosts the Signal & Noise podcast and has spent 15+ years specializing in online panel quality and B2B sample. You can catch the full webinar replay at roirocket.com.
