Myth-odologies: The truth behind our market intelligence

Every industry has its assumptions. Market research is no different. 

In this special Euromonitor audio feature, you’ll hear from Gayatri Bhasin Darke, Director of Research Transformation, who breaks down seven myths about research, data science, forecasting and AI—unpacking the reality behind these common beliefs. Drawing on years of experience and practical examples, she explains how insights are built, how forecasts are developed and why context matters as much as the numbers themselves.  

Whether you work in strategy, innovation, consumer insights, market intelligence or consulting, this episode offers a behind-the-scenes look at the methods, specialist expertise and human judgment that turn information into better decisions. 

To learn about our market intelligence, visit euromonitor.com.  

For more information on our research methodology, visit euromonitor.com/who-we-are/research-methodology.  

00:00:39 Myth 1: The bigger the sample, the better the data 

00:02:51 Myth 2: One good data set is all you need 

00:05:07 Myth 3: Categories mean the same thing everywhere 

00:07:20 Myth 4: Reliable data is 100% accurate 

00:09:23 Myth 5: A forecast is just an educated guess 

00:11:38 Myth 6: The data speaks for itself 

00:13:07 Myth 7: AI can do this already 

Lots of brands claim to be number one… but can they prove it?

At Euromonitor International, we help brands build trust through evidence-based research. Our claim validation service ensures your marketing messages are backed by real data. Stand out in a crowded market. 

Visit euromonitor.com/claims to learn more.

Introducing Opportunity Minded, a new series from Euromonitor International designed for forward-thinking business leaders like you.  Each episode tackles a strategic approach or topic on corporate agendas.  You’ll hear from our experts who share in

Transcript

0:23

Every industry has its assumptions and market intelligence is no different. In this Euromonitor podcast, you'll hear seven myths about research, data science, forecasting and AI, drawing on the perspectives and expertise of Gayatri Bhasin Darke, our Director of Research Transformation. Some of these assumptions may sound familiar. Some may feel completely reasonable, but each one opens the door to a bigger conversation about how insights are built, how forecasts are developed and how we can help you make sense of it all. Let's start with a common assumption.“The bigger the sample, the better the data.” At first glance, this feels obviously true. More inputs, more confidence. A bigger sample must be a better one. It's how most of us are taught to think about evidence. The larger the survey, the more you can trust it. And in plenty of context, that instinct is exactly right Where it breaks down is the assumption that every conversation is worth the same. Behind this myth sits some really reasonable client questions. Is a handful of interviews really enough to shape a whole industry? How do you decide who to talk to? What stops one loud voice from skewing the result? The honest answer is that we're not running a random poll. We deliberately choose the people who can see the whole market. Manufacturers, distributors, retailers, trade bodies, and who we approach as a considered decision. Can a trade body give us unpublished data? Will the market leaders engage whether they're clients or not? If not, which emerging brands can still speak credibly to how consumers are behaving and how the competitive landscape is shifting. This isn't cold calling, we’re nurturing contacts year on year, speaking to them sometimes several times a year. So the reality is that a few of the right conversations, weighted and triangulated against big data, beat a thousand of the wrong ones. Doubling a survey sample barely moves accuracy. Talking to the one distributor who can see the whole market universe can change the picture entirely. And it isn't only who we talk to. It's where they sit. Our analysts are in-country. They speak the language, they have the relationships and they have the cultural context that catches the outlier that's actually real. When ice cream sales in Mexico looked plainly wrong from a head office view. It took a local analyst and an in-depth trade conversation to explain the swing In sportswear, a now annual dialogue with one global brands UK and regional teams has sharpened our market sizes and even helped us refine our menswear versus women's wear split. So judge research by the quality and fit of its inputs, not the raw count. A few of the right conversations, conducted by people who understand the market tell you more than volume ever could. If better research is not simply about doing more, that could lead to our next assumption.“One good data set is all you need.” It's an appealing idea. Find the one authoritative source and you're done. It feels efficient. And there's something reassuring about imagining a single clean number sitting behind a whole market. Most of us would love research to be that simple. The questions behind this one are familiar. Where does the market number actually come from? What are your sources? And isn't it all just desk research? Let's clear that last one up first. Desk research isn't typing a question into a search engine. It's structured analysis of official statistics, trade body data and company filings. Alongside it, we draw on big data, clickstream receipts, web-scraped product listings and our own primary research in the field. The deeper point is that no single data set sees the whole picture. Each one captures something the others miss and each carries its own blind spots. So we built every number from many independent sources and reconcile them. Two routes meet in the middle. National statistics and company reporting set the outer boundary a market can't exceed. Store checks, scraped prices and real online purchases help us build the demand picture. And where they disagree, our analysts reconcile them. Often a single trade conversation or a client insights exchange helps explain that gap. We're also pragmatic . Where we can't reliably gather something ourselves, we buy it. Chinese e-commerce data, for example, is notoriously hard to scrape well, so we pay specialists for it rather than pretend otherwise. And a lot of real demand never crosses the till. Informal stalls, independent retail, illicit trade. In tobacco, we use duty data, conversion factors and expert estimation to size the illicit segment that no scanner ever captures. We’re measuring actual demand, not just the transactions that happen to be recorded. Treat any single source with healthy caution.

Real confidence comes from three things: 5:05

triangulating many independent signals, expert judgement to know which of those signals to trust and taking the numbers out to the trade, to the market, to test them. Building a market view usually means bringing together information from many different sources, but collecting data is only half the challenge. Is your data comparable? You might assume- “A category means the same thing in every market.” This one rarely gets questioned because it seems too obvious to doubt. Yoghurt is yoghurt. A supermarket is a supermarket. When you're comparing markets, you naturally assume you're comparing like with like. The alternative barely occurs to you. But the question hiding here is a serious one. How can you compare category across 100 markets and 20 years and be sure it isn’t apples and oranges? In one country, yoghurt may include drinking yoghurt. In another, it doesn't. One market’s convenience store is and other’s grocery retailer. Where should we cover karaoke within consumer food service. This is where our taxonomy earns its keep. A single unified way of defining every category, channel and market refined over 50 years. It sounds unglamorous, but it's actually the thing that makes global comparison possible. Because we hold every market to the same definitions. You can line a category up across countries, track it consistently over time and know the movement you're seeing is real and not a measurement artefact. Keeping it clean at that scale takes both machine and human effort. So automated checks flag outliers and inconsistencies. Then analysts stress test it against what they know on the ground. When we rebuilt our ingredients data, we combined millions of scraped product listings, nutrition information, pricing, ingredient information, with analyst expertise, all mapped back to one taxonomy, turning a mountain of inconsistent listings into something you can genuinely compare. Or if we really want to go into the weeds on this. In one market formula milk aimed at 1 to 2 year olds sits under standard milk formula. In another its follow-on milk. We standardise the definition around the age range the product is actually marketed at and then hold every market to it. Apples to apples Comparability isn't a given. It’s engineered. Consistent definitions are what let you make a confident decision across markets, which is exactly when a hidden mismatch could cost you the most. Of course, even when data is collected consistently, people still want to know one thing. Can I trust it? Let's talk about that. “Well. Reliable data is 100% accurate.” There's something reassuring about this idea. If you're paying for data, it should be right. Full stop. Precision feels like the whole point, and nobody enjoys being told a number carries uncertainty. It's a completely understandable thing to want.

The real questions underneath are: 8:33

how accurate is this really? when my own number doesn't match yours, who's wrong? and, how do you validate any of it? Here's the uncomfortable truth we’d rather say out loud than hide. No provider can be 100% accurate, and anyone who claims to be should make you a little nervous. I think given where we are in the world now, it's increasingly important to acknowledge uncertainty and learn to sit with it at times. What we can tell you is where to lean hard and where to lean lightly. Confidence is highest at established top line categories, where many sources converge and more measured in small, fragmented or fast moving segments. That's normal, and being open about it is a strength and not a weakness. Before anything is published, automated checks flag inconsistencies. Our experts stress test the data and clients validate their own representation in it. When your internal number differs from ours, it's almost never a fight. It's usually a difference of definition or scope, and working through it improves the data for everyone. One partnership with a global consumer goods manufacturer grew from us simply delivering research into genuinely solving problems together, with regular validation checkpoints and wider engagement across their teams. Over time, they opened validation across dozens of markets, many of them data-poor countries, and that input materially improved the picture. None of this happens in a one directional process. Demand honesty about confidence, not forced precision. Knowing where a number is rock solid and where it's directional is what lets you weight it correctly in a decision. Understanding what happened is valuable. Understanding what might happen next is often even more valuable.

But many people think: 10:17

“A forecast is just an educated guess.” It's easy to see why this comes from. Forecasts are about the future. The future is uncertain and we've all watched confident predictions age badly. Calling a forecast a guess can feel like healthy scepticism rather than cynicism. The questions behind it are fair. What actually goes into a forecast? How do you handle something that depends on a regulation that isn't final yet? And when a shock hits– a tariff, a conflict, –how fast can you really react? The starting point is that a forecast isn't a crystal ball. Is the model of the things that genuinely move a market. We begin with the hard drivers– income, GDP per capita, price, demographics, urbanisation –and estimate statistically how sensitive each category is to them, using decades of history, price and income elasticities and borrowing strength from similar markets so a smaller country isn't modelled in isolation. This produces several statistically sound models. Then senior industry experts choose the one that makes the most business sense, not just the neatest fit. Finally, local analysts adjust for what no equation captures: a regulation coming in, a channel shift, consumer sentiment. The result is a consensus baseline deliberately measured with every assumption written down. You can picture it as two stages. A first statistical stage produces the baseline, and a second stage puts it in front of local and regional specialists to review, sense, check and refine before anything is published. Machine speed, then human judgement. When a shock occurs, the response is not simply to adjust the forecast number, there is necessarily a lag. The team updates baseline assumptions, develops multiple scenarios around the event and combines modelling with expert judgement to assess the impact on growth, inflation and consumer demand. This helps our clients understand a range of possible outcomes and make more resilient decisions in times of uncertainty. A good forecast isn't a promise. It's a defensible base case., plus the levers to test your own what-ifs. For planning, that's far more useful than false certainty. Forecasts help inform decisions, but decisions need more than information alone. And that's where our next assumption comes in: “The data speaks for itself.” This sounds almost virtuous. Let the numbers talk. Don't editorialise In a world of spin and fake news,“Just give me the data” feels rigorous and objective. But the questions underneath tell a different story. You've given me the data., now what do I do with it? What's the so-what? Is this a blip or a real shift? A number on its own doesn't make a decision. And two capable people can read the same data and reach opposite conclusions. Our job doesn't end when the data lands in Passport or in your inbox. The value is also in the interpretation. What's driving this? Where is it heading and what you should do about it? That's where our experts come in. Connecting a number to the forces around it and translating it into something you can act on. To give you a recent example, a Korean skincare brand engaged our Seoul team to understand the UK skincare market and shape a potential channel expansion move. Our UK beauty experts translated global K-beauty strengths into concrete UK opportunities– longevity, science led skincare and a credible local positioning –giving them a clear sense of where to play and a foundation to widen the engagement. The raw number is the start, not the end. What turns information into an advantage is the context and the interpretation that tell you what to do next. Finally, we arrive at the query that's reshaping almost every conversation about knowledge, expertise, and information.“AI can already do all of this.” This is the assumption on everyone's mind right now, and for good reason. AI tools are genuinely impressive. They summarise, they synthesise. They answer instantly. If a model can read the whole internet. It's completely fair to ask why you'd pay for research at all. The questions behind it are pointed. Why can't an AI tool just do this? Do you use AI to make up your numbers? What do your analysts know that isn't online anyways? Let me concede the honest part first. A general model can summarise what's already public and so can we, faster. In fact, we've used AI and machine learning for years to do exactly that. Our data science teams already use AI heavily to scrape, classify and link the vast, messy streams of data that feed every number. At a scale no team could match manually. So AI does enormous work inside our process. But here's a line it can’t cross. A model can't visit a store in Lagos, interview a distributor in Sao Paulo, or persuade a manufacturer to share confidential volumes it has never published anywhere. That proprietary, locally-gathered, human-validated layer is what no model can fabricate, and our taxonomy is what makes it comparable across markets. So to be precise, we use AI to process the data, not to invent the numbers. The judgement about which sources to trust stays with our people. Building market shares and sizes is judgement heavy. Knowing which sources to trust, which numbers to challenge, which outputs need a second look. The real question isn't AI versus human, it's who uses AI well. As everyone leans on these tools, the proprietary local validated layer gets even more valuable. We've covered a lot of ground, from expert interviews and data sources to forecasting, decision making and AI. And if there's one theme connecting all these topics, it's this: the most useful answers are rarely the simplest ones. Behind every number is context. Behind every forecast is expertise. And behind every insight is a combination of evidence, human judgement and years of experience. The assumptions explored today aren't unreasonable. They're often the starting point for some of the best conversations. And this conversation doesn't have to stop here. Visit the links in the description to learn more about Euromonitor’s mixed methodology approach and connect with our experts.