IB Business Sampling Methods In Market Research

When 190,000 People Sampling Tell You You're Right... But You're Still Wrong

23rd April 1985. Roberto Goizueta, CEO of Coca-Cola, stood before 200 reporters at New York's Lincoln Center with the confidence of someone who plays poker and holds a royal flush. After conducting nearly 200,000 taste tests and spending $4 million on market research, he was about to announce the biggest product change in Coca-Cola's 99-year history. The numbers were "staggeringly superior," he beamed. Science itself had "blessed" this decision.

Three months later, the company was bringing back the original formula after one of the most spectacular marketing disasters in history. Angry customers had flooded customer service phone lines with 8,000 calls a day. People were hoarding old bottles like they were preparing for an apocalypse. One protestor compared it to a violation of freedom as fundamental as the Magna Carta. Peter Jennings interrupted the TV soap opera General Hospital to announce that old Coke was coming back - that's how big a deal this was.

So what went wrong? How could 190,000 people be "right" and yet everything go so wrong?

The answer lies in one of the most crucial concepts in business: sampling.

Why It's No Good Asking Everyone Everything

Right, let's clarify some basic IB Business Management concepts. In market research, population refers to everyone who could potentially be a customer for your product. For Coca-Cola in 1985, that was basically every human being in America who wasn't living under a rock.

Now imagine trying to ask all 240 million Americans whether they liked New Coke. You'd need to find them, talk to them, record their answers, analyse the data... and by the time you finished, everyone would have moved on to whatever drink TikTok was promoting. (TikTok didn't exist. You get the point.)

This is where sampling saves the day. It's the practice of selecting a smaller group - a sample - from that massive population to represent everyone else's views. Done right, you can ask 1,000 people what they think and get a pretty accurate picture of what 240 million people think. Done wrong, and you end up pouring millions of litres of unwanted cola down the drain.

Three things matter when sampling:

1. It has to be representative - your small group needs to reflect the bigger population

2. It's practical - asking everyone is impossible, too expensive, and unnecessary

3. It needs to be done properly - which brings us to the three main methods

The Three Ways to Pick Your Sample

Quota Sampling: When You Need Specific People for Specific Reasons

Think about Spotify for a second. Right now, they know that 30% of their users are aged 18-24, and another 31% are 25-34. These aren't just stats - they're absolute gold for market research. If Spotify wants to test a new feature, they can't just ask any random people off the street. They need to mirror their actual audience.

This is quota sampling in action. You predetermine how many people from each sub-group you need. So if you're surveying 100 people for Spotify, you'd grab 30 from the 18-24 age bracket, 31 from the 25-34 crowd, and split the rest according to your actual user demographics.

The quota method works brilliantly when:

• You're investigating specific characteristics (like age, gender, income)

• You're comparing different sub-groups

• You need your sample to match your actual market proportions

IB Business Management Real-life Example: Take Shein, which absolutely dominates the Gen Z fast fashion market with a 50% market share in the US. When they research new clothing lines, they're not randomly asking 60-year-old accountants what they think. They're using quota sampling to ensure they're talking to enough Gen Z women (their primary market), enough international customers, enough mobile-first shoppers. They've identified their segments and they sample accordingly.

Issues with the approach? You need to know your market characteristics in advance. If you guess wrong about what matters - say, you sample based on age when the real determining factor is income - your research will look great and be completely useless.

Random Sampling: The Democratic Approach That Sounds Perfect But Isn't

Random sampling is exactly what it sounds like: everyone in the population has an equal chance of being selected. It's usually done by computer from a database, which means it's quick, cheap, and brilliantly unbiased.

In theory, it's the fairest method. No human prejudice. No "I'll just ask people who look friendly." Just pure mathematical democracy.

The problem? Pure randomness can produce extremely unrepresentative results.

Imagine you're researching customer satisfaction for a gym chain. You randomly select 100 members from your database. By sheer chance, you might end up with 70 people who joined in January (New Year's resolution crowd) and only 10 who've been members for years. Your "random" sample would wildly overrepresent beginners and completely miss the perspective of loyal long-term members.

This is why random sampling works best when:

• Your population is fairly homogeneous (similar)

• You need results quickly

• Cost is a major factor

• You're not too fussed about precise sub-group analysis

But if you're dealing with diverse populations - like, say, everyone who drinks Coca-Cola - random sampling can be a bit like playing Russian roulette with your research budget.

Convenience Sampling: "We'll Ask Whoever's Around"

Convenience sampling is the method you use when you're standing in a shopping centre with a clipboard, asking people, "Got a minute?" It's sampling whoever happens to be easily accessible at that moment.

Universities love this method because students are always available. ("Would first-year psychology students like to participate in a study? 100% of them said yes because they need the extra credit.")

The appeal is obvious: it's fast, it's cheap, and subjects literally come to you.

But it could be a total disaster... Your findings could be so unbalanced and biased they're basically fiction.

A recent market research analysis from 2024 highlighted this perfectly: if you conduct a survey about product preferences using only your existing customers (convenience!), you might completely miss the views of potential new customers or untapped demographics. You end up with research that tells you what people who already like you think... which is a bit like asking your mum if you're handsome / pretty.

The method works for preliminary research or when you're just exploring ideas. But if you're making major decisions based on convenience sampling - like, oh I don't know, changing a 99-year-old formula - you're essentially gambling your company's reputation on whoever happened to be free that day.

Back to New Coke: The Day Brilliant Research Became a Billion-Dollar Howler

Now we can properly understand what went wrong with New Coke.

Coca-Cola did conduct 190,000 taste tests. That's not a small number - that's larger than the population of Cambridge. The tests used a combination of random sampling (to avoid bias) and elements of convenience sampling (taste tests in shopping centres and public spaces where people were accessible).

But it all fell apart:

1. The Tests Were "Sip Tests" People tasted tiny samples - just a few sips. Malcolm Gladwell later pointed out in his book Blink that this created a systematic bias toward sweeter drinks. In small amounts, sweeter tastes nice. Over a whole can? It becomes cloying. The sampling method itself was flawed because it didn't represent actual consumption behaviour.

2. They Asked the Wrong Questions Researchers never asked, "Would you be happy if we replaced original Coke with this new formula?" They just asked if people liked the new taste. That's a sampling error - the mistake arose from how they designed the research itself.

3. They Ignored the Vocal Minority About 11% of taste-testers were furious at the idea of changing Coke, even if they preferred the new taste in blind tests. Coca-Cola dismissed this as noise. Turns out, that 11% was loud enough to shape public opinion and trigger a nationwide revolt. They failed to recognise that in business, a substantial minority isn't just "people who disagree" - they're revenue streams you're about to lose and worse, a minority with the ability to influence a substantial majority.

4. They Completely Missed Emotional Attachment The research focused entirely on taste (quantitative data from a specific sample) but ignored the fact that Coca-Cola wasn't just a drink - it was American culture in a bottle. That's a non-sampling error: a mistake that wasn't about who they asked or how many, but about asking the wrong thing entirely.

As Coca-Cola's president later admitted: "The simple fact is that all the time and money and skill poured into consumer research could not measure or reveal the deep and abiding emotional attachment to original Coca-Cola felt by so many people."

When Sampling Goes Right:

IB Business Management Real-life Examples: Who is getting sampling right in 2024-2025?

Spotify uses sophisticated quota sampling combined with big data. They don't just randomly survey users - they carefully segment by age (18-24, 25-34, 35+), listening habits (passive vs. active users), premium vs. free tier, podcast listeners vs. music-only... and they sample proportionally from each segment. This is why their Discover Weekly playlists feel eerily accurate. They're not guessing; they're sampling the right people and asking the right questions.

Netflix famously stopped sharing subscriber numbers in 2025, but when they were conducting research, they used massive random sampling backed by viewing data. Because their entire user base was digitally tracked, they could randomly sample millions of viewing sessions and still get statistically significant results. That's random sampling at scale.

Shein targets Gen Z with laser precision. They know that 30% of US and UK Gen Z consumers shop with them, and their primary demographic is women aged 18-24. Their market research uses quota sampling that mirrors these demographics - they're not asking 50-year-old men what they think of their latest crop tops. They sample based on their actual market segments and test thousands of styles weekly, using sales data as constant market research feedback.

The Two Types of Errors That Can Destroy Your Research Efforts

Even with perfect sampling methods, things can still go wrong. Market researchers distinguish between two types of errors:

Sampling Errors are mistakes that come from your sampling design:

• Sample size too small (asking 10 people to represent 10 million)

• Unrepresentative sample (asking only university students about retirement planning)

• Wrong sampling method (using convenience when you needed quota)

• Built-in bias (only surveying people who already like your product)

A 2024 analysis from Cognitive Market Research found that small sample sizes are one of the most common issues - they lead to conclusions heavily influenced by outliers rather than actual market trends. If your sample's too small, you're basically reading tea leaves.

Non-Sampling Errors are mistakes not related to who you asked:

• Respondents lying or giving socially acceptable answers instead of honest ones

• Questions worded in confusing or leading ways

• Data entry mistakes

• Misinterpreting correlations as causation

Here's a classic non-sampling error from 2024 market research: a company saw sales spike after a marketing campaign and attributed it entirely to the campaign. What they missed? The spike coincided with the Christmas shopping season and a competitor going out of business. The error wasn't in who they sampled - it was in how they interpreted the data.

Best Ways to Present Sampling

Once you've collected all this data, you need to present it in ways that don't make people's eyes glaze over. The main formats are:

Pie charts for percentages - "45% of respondents aged 18-24 prefer TikTok over Instagram"

Line graphs for trends over time - like Netflix's subscriber growth from 269.6 million in Q4 2023 to approximately 301.6 million by 2024

Bar charts for comparing frequencies - Shein's market share growth from 18% in March 2020 to 50% by November 2022 looks dramatic in a bar chart

Tables for detailed numerical data - especially when you're comparing multiple variables across different segments

For qualitative results (opinions, feelings, explanations), you'd summarise the main themes in text, pull out representative quotes, or use infographics to visualise common responses.

IB Business Management exam tip: When you're analysing charts and graphs in your IB exams, always check the axes carefully. Don't assume the examiner knows what you understand. A graph showing "sales growth" could mean total sales, sales percentage increase, or sales compared to last year - each telling a completely different story.

IB Business Management Exam Corner

What separates brilliant market research from expensive disasters?

1. Match your method to your question

• Need precise sub-group comparisons? Quota sampling.

• Want quick, unbiased results from a uniform population? Random sampling.

• Just exploring ideas or doing preliminary research? Convenience sampling might be fine.

2. Recognise your method's limitations New Coke failed partly because they used convenience sampling (taste tests in public places) and treated the results as if they'd used rigorous quota sampling that captured emotional attachment. Know what your method can't tell you.

3. Sample size matters - but so does quality 190,000 badly-designed taste tests couldn't prevent disaster. Sometimes 1,000 well-designed interviews will give you better insights than 100,000 rushed surveys.

4. Remember: samples can't capture everything No matter how good your sampling, you're always working with incomplete information. Stay humble. Test your assumptions. And if 11% of people are screaming that you're about to make a terrible mistake, maybe listen to them.

The fast fashion industry are very efficient at this. Companies like Shein sample constantly - they release thousands of styles in small batches (essentially using sales data as continuous sampling) and only mass-produce what sells. They've turned their entire business model into one continuous market research exercise. That's quota sampling meeting real-time feedback, and it's why they've captured half the US fast fashion market.

Final Thought For Your IB Business Management Course

In the end, Coca-Cola learned what researchers have known for decades but companies keep forgetting: you're not sampling data points. You're sampling human beings with memories, emotions, and surprisingly strong opinions about soft drinks.

The population you're researching isn't a spreadsheet. It's people - contradictory, emotional people who might say they prefer the sweeter formula in a taste test but will stage protests when you try to take away the drink they grew up with.

Get your sampling right, and you unlock genuine insights into what people want. Get it wrong, and you'll join New Coke in the marketing hall of shame - a tale told to business students for generations.

Stay well,