What does the insights industry actually struggle with? Notes from 19 recorded conversations in 2026

What does the insights industry actually struggle with?
Notes from 19 recorded conversations in 2026
Written by: Mark Nguyen, ND Research
Originally Published: October 23, 2026
Labels: Field Notes, AI-assisted, Continuously Maintained
Topics: Consumer insights, Brand strategy, Market research, Social listening, Synthetic data
TLDR: We recorded 19 conversations across the insights world this year. Nobody was short of data. The most common complaint (9 of 19) was that research doesn't connect to a decision.
Key Ideas:
- Across 19 recorded industry conversations (April to October 2026), the most common ask was insight tied to an action or a KPI. It came up in 9 of the 19. "Interesting" no longer clears the bar.
- 6 of 19 said feeds have become so personalized that today's research tools miss what people actually see.
- 6 of 19 raised budget pressure. Research is often one of the first lines to get cut.
- 6 of 19 were skeptical of synthetic respondents or AI hype, including the two synthetic-data vendors we spoke with, who set firm limits on their own products.
- Nobody outside our team brought up survey fatigue, though in fairness we never asked about it directly.
Preamble
We started recording our industry conversations this spring mostly to keep ourselves honest. Notes drift; transcripts don't. Six months later we had 19 of them, and reading them back to back felt a bit like reading the same letter from 17 different people. This is the tally, with the numbers attached so you can argue with them.
Obviously it's a small sample, and it leans toward people willing to take a call from a startup. At the same time, it's the closest thing we have to real evidence about what the insights world is wrestling with right now, so I'd rather publish it with its limits than sit on it.
What problems came up in 19 conversations about strategy and insights work?
Short answer: nobody is short of data. Most of what they get just doesn't change a decision.
Between April 1 and October 20, 2026, we recorded 19 conversations with 17 people and organizations across the industry: agency and holding-company leaders, a brand-side creative lead, research and measurement veterans, independent advisors and consultants, and four vendors in social listening and synthetic data [1]. Just for full transparency, none of them held the title of brand strategist or planner.
These are the people who sell to strategists, advise them, or run the research they use. We went through the transcripts and counted how many separate conversations raised each theme. Here's how it came out:

- Insight has to lead to an action or a KPI: 9 of 19
- Personalized feeds that current tools can't see into: 6 of 19
- The context around an ad matters for whether it lands: 6 of 19
- Budget pressure on research: 6 of 19
- Skepticism of synthetic respondents or AI hype: 6 of 19
- Our own feed data was hard to grasp at first: 5 of 19
- Small samples and base sizes: 5 of 19
- Proof first (a case study or an A/B test): 4 of 19
- Privacy, consent and personal data: 4 of 19
- Social listening gives volume without context: 3 of 19
- Linking what people saw to what they bought: 3 of 19
- Survey fatigue: 0 of 19
It's a small and mixed group, to be frank. More on how we counted (and what that means) at the end.
9 of 19 - conversations in which insight had to connect to an action or a KPI before it counted
Why isn't an "interesting" insight enough anymore?
Because the people who pay for research have to defend it in a budget meeting, and "interesting" rarely survives that meeting.
This was the most consistent thing we heard. A senior executive at a large advertising holding company asked early on which KPI our data would move. A long-time CPG and retail measurement veteran told us about a transaction dashboard his team built: people found it interesting but kept asking what to do with it.
It only clicked once they turned it into a site-selection tool that said, in effect, put your next store in this plaza. His view: the further you move from insight toward a recommendation, the easier the sale, and the more work it takes to deliver.
Two advisors, a research-agency founder and that same CPG veteran, pushed us the same way about three weeks apart: move from describing what people see to recommending what to do about it (e.g. which creators or placements). Two friendly voices aren't proof, but they line up with the 9 of 19 who asked which KPI our data would move.
The meta lesson for anyone selling research (us included): price the decision your chart makes easier, not the chart.
Can social listening or panels show what audiences actually see?
Not at the level of one person's feed. Interestingly, some of the clearest answers came from people who build and sell these tools.
Social listening works top-down. You search for a keyword, a hashtag or a topic, and you get back the volume and sentiment of public posts. That's useful for share of voice or a crisis alert.
In 3 of our 19 conversations, people described its limits: keyword-level, aggregate and, in one executive's words, numbers without context. (One listening vendor pushed back, saying Boolean search can tie conversations to demographics.) An account executive at a media-monitoring platform told us, on a sales call, that their tool works at the level of keywords and media audits and doesn't get down to an individual consumer's habits.
Device metering has a different blind spot. A measurement leader at a global research firm explained that their meters can tell which social apps a person opens, but not what those apps show them. The social feed is a known gap for meters like theirs.
Underneath both is the shift that came up in 6 of our 19 conversations: feeds are personalized, so two people on the same app are basically on different internets. A CPG measurement veteran guessed he'd scrolled past about 200 clips the night before our call, with ads, creators and news all mixed together.
Very few people outside the platforms see that mix person by person, and the tools that try mostly capture the ads, not the feed around them. That's the gap we've been working on (we explain the category in What is feed exposure intelligence? [2]).
Is research budget being cut?
For a lot of teams, yes, or at least it's under much tighter scrutiny. Budget pressure came up in 6 of 19 conversations.
An agency-side marketing executive described research as something of a luxury: it can get cut, and it competes with other research budgets. The founder of a US research agency told us budgets were shrinking. Others talked about squeezed agency margins, and about consolidation and layoffs at research firms.
Advisors pointed to two things that tend to survive that pressure:
- Research framed as efficiency, not as more insight. Think: same budget, more of the right people reached, or a lower cost to acquire a customer.
- Niches with their own budget. Youth and Gen Z insight came up twice (one advisor's point, paraphrased: senior marketers don't know how to reach Gen Z, so they'll pay to find out). Political research came up three times.
Is synthetic data replacing real respondents?
Not in the conversations we had. 4 of 19 were skeptical of synthetic respondents specifically, and 2 more were wary of AI hype in general.
The most striking version came from a research-agency founder whose firm has invested in AI for about two years, including synthetic data and digital twins. Clients were curious, he said, but few were buying. The measurement leader at the global research firm drew the line at synthetic respondents with nothing real underneath, while still using synthetic modelling built on a real metered panel.
The two synthetic-data vendors we spoke with were among the most careful voices of all. We went deeper on this in Will synthetic respondents replace real consumer research? [3]
Do you need a census-representative sample?
For a lot of brand questions, perhaps not. What matters more is having enough people in the segment you actually plan to act on.
Sample size came up in 5 of 19 conversations. The sharpest advice came from an experienced research-agency founder: stop chasing census representation and build up base sizes for the audiences clients will pay for, roughly 150 to 300 per segment.
Almost all of his work is segment-based (think: Gen Z or millennials). He was also critical of panel platforms that project a small sample onto a universe of millions, for segments that aren't nearly that big.
A measurement-company founder made a similar point about two weeks later: start lean, with sample or public data, before paying for a large representative sample. Honestly, we're still figuring out what this means for our own panel design.
What makes a new kind of data a must-have instead of a nice-to-have?
From what we heard, it has to clear a few bars:
- A one-sentence explanation. In 5 of 19 conversations, people found our own feed data hard to grasp at first (too high level came up more than once).
- Proof. 4 of 19 asked for a case study, an A/B test, or evidence that a pattern is statistically meaningful and not a fluke.
- A KPI that matches the buyer. One advisor's framing: media efficiency for agencies, incremental revenue for research firms, sales lift for brands.
Just for full transparency, that first one was about us: all five of those reactions were to Tapestry. Feed data is a new category, and in hindsight we've had to learn to lead with the decision it supports rather than with the data itself.
What didn't come up?
Survey fatigue. We expected to hear about it, since response rates are a constant worry inside research operations (ours included). But across 19 conversations, not one person outside our team raised it. We didn't ask about it directly, so treat this as a gap in what came up, not proof that it isn't a problem.
At the same time, I'd wager it says something about where the pain actually sits: the complaints we heard were about what research produces, not how hard it is to field.
How we gathered this
- What: 19 recorded conversations with 17 distinct people and organizations, April 1 to October 20, 2026 [1].
- Who: agency and holding-company leaders (4 conversations, 3 organizations), a brand-side creative lead (1), research and measurement veterans (5 conversations, 4 people), independent advisors and consultants in marketing, ad tech and AI (5), and vendors in social listening and synthetic data (4).
- How we counted: a theme counts once per conversation, and only when someone outside ND Research raised it or clearly agreed with it. We checked counts against full transcripts for 18 of the 19 conversations (one had only a summary).
- Limits: this is a small, mixed group, not a survey. Only 5 of the 19 conversations were with potential buyers, and none of the people held the title of brand strategist or planner. Several are advisors to us, so they aren't neutral. Treat the counts as directional.
- Privacy: we describe people by role and organization type only, and we paraphrase apart from a few short phrases, because these were private conversations.
Appendix: the 19 conversations
For anyone who wants to weigh the counts above, here is every conversation we counted, by month, role and how it came about. People and organizations are described by type only. Themes are the ones we tallied for that conversation, so you can check our arithmetic.
- April 2026. Senior innovation and insights executive at a large advertising and marketing holding company, with a colleague. An intro call that became a pilot proposal. Themes: insight tied to a KPI, personalized feeds, proof, limits of social listening, cost of platform data, hard to grasp at first.
- April 2026. Inventor and founder with ties to music and creative agencies. An advisor intro. Summary only, little usable signal.
- April 2026. Independent consultant and psychologist who diagnoses AI deployments. Expert conversation. Themes: personalized feeds, privacy and consent.
- April 2026. Sales representative at a social listening platform. A sales call we took to evaluate the tool. Themes: what listening covers, and a counterpoint that Boolean search can tie conversations to demographics.
- April 2026. Marketing executive and consultant with an agency ad-operations background. Expert feedback after a product demo. Themes: insight tied to a KPI, context around the ad, budget pressure, AI hype, hard to grasp at first, proof, political research, localization, speed.
- May 2026. Chief product officer at a synthetic-data startup. Vendor conversation. Themes: limits of synthetic respondents, base sizes.
- June 2026. Head of investments at the same holding company as the first conversation, with two colleagues. A strategic-fit conversation. Themes: privacy, cost of platform data, whether platforms already do this.
- June 2026. Owner of a small marketing agency who was also a beta participant. Pilot and product feedback. Themes: insight tied to a KPI, personalized feeds, context around the ad, budget pressure, privacy, political research, localization.
- July 2026. Founder of an early-stage ad-tech startup with an agency background. Expert conversation and introductions. Themes: context around the ad, budget pressure, who the buyer is.
- July 2026. Senior executive at an influencer-marketing company with a long agency career. A pilot idea. Themes: insight tied to a KPI, context around the ad, hard to grasp at first, sample size, proof, attribution, how creators get picked.
- July 2026. Head of an in-house creative studio at a large technology company. Brand-side prospect call. Themes: personalized feeds, speed of vendors.
- August 2026. Founder and CEO of a US market-research agency. First advisor call. Themes: insight tied to a KPI, context around the ad, budget pressure, hard to grasp at first, political research.
- August 2026. Agency-side marketing executive and advisor. Advisor call. Themes: insight tied to a KPI, budget pressure, proof, Gen Z insight.
- September 2026. Senior audience-measurement leader at a global market research firm. Research-partner conversation. Themes: insight tied to a KPI, personalized feeds, context around the ad, synthetic respondents, sample size, privacy, limits of metering and listening.
- September 2026. Founder of a synthetic-survey startup. Vendor conversation. Themes: limits of synthetic respondents, sample size, refusing to synthesize feed content.
- September 2026. The same research-agency founder as in August. A live review of the product. Themes: insight tied to a KPI, synthetic respondents, hard to grasp at first, base sizes, Gen Z insight, whether platforms already do this.
- October 2026. Founder of a media-measurement company. Advisor call. Themes: attribution, start lean, a zeitgeist read for media and entertainment.
- October 2026. CPG and retail measurement veteran, now independent. Expert conversation. Themes: insight tied to a KPI, personalized feeds, budget pressure, attribution, platform data locked away, the say-do gap.
- October 2026. Account executive at a media-monitoring platform. A sales call we took to evaluate the tool. Themes: limits of social listening.
Two organizations appear twice (the holding company and the research agency), which is how 19 conversations come to 17 distinct people and organizations. Five of the 19 were with potential buyers. None were with a brand strategist or planner by title.
FAQ
What is the biggest problem in consumer insights in 2026?
In our 19 conversations (none with a strategist or planner by title), the most common theme was turning insight into action: 9 of 19 said research has to connect to a decision or a KPI, and interesting isn't enough to keep a budget.
Why can't social listening show what consumers see in their feeds?
Social listening searches public posts by keyword or topic, so it shows what people choose to say in public, in aggregate. It doesn't show the personalized mix of creator posts, memes and ads that an algorithm serves one specific person.
Are brands buying synthetic respondents?
Interest is high and buying is cautious. 4 of our 19 conversations were skeptical of synthetic respondents, and a research-agency founder whose firm has invested in them told us clients were curious but few were buying.
How big should a sample be for segment-level consumer insight?
One experienced research-agency founder's rule of thumb: roughly 150 to 300 respondents in each segment you plan to act on, rather than a census-balanced sample where each segment ends up thin.
How to cite this
Nguyen, M. (2026, October 23). What does the insights industry actually struggle with? Notes from 19 recorded conversations in 2026. ND Research. https://www.nguyendigital.co/blog-posts/what-strategists-and-insights-teams-struggle-with
List of References
(Continuously Maintained)
[1] ND Research (2026). Industry conversation log: 19 recorded conversations with 17 people and organizations, April 1 to October 20, 2026. Agency and holding-company leaders (4), a brand-side creative lead (1), research and measurement professionals (5), independent advisors and consultants (5), and social listening and synthetic-data vendors (4). Counts checked against full transcripts for 18 of 19. Each conversation is listed by month, role, how it came about and the themes it raised in the appendix above. Participants are described by role only.
[2] ND Research. Nguyen, M. (2026, October 23). What is feed exposure intelligence? The missing layer in consumer research. https://www.nguyendigital.co/blog-posts/what-is-feed-exposure-intelligence
[3] ND Research. Nguyen, M. (2026, October 23). Will synthetic respondents replace real consumer research? What researchers and vendors told us. https://www.nguyendigital.co/blog-posts/will-synthetic-respondents-replace-consumer-research





