Survey Sample Example: See How to Define Yours

If you've already started putting together a survey, whether for work, college, or to better understand your company's customers, you've probably run into a question that stumps many people: who, after all, should answer your questionnaire? This question has a name: survey sample. And understanding this concept well, with a survey sample example in hand, is what separates a reliable study from one full of flaws.

In this article, we will discuss what a sample is, why it matters so much, what types exist, how to calculate the ideal size, and, of course, look at practical examples you can apply to your own survey. The idea is that, by the end of the reading, you will know exactly how to choose your sample without relying on guesswork.

What is a survey sample?

A survey sample is the group of people (or items, depending on what you are studying) selected from a larger universe, called the population, to represent that universe in a study. Instead of interviewing everyone you would like to understand, you interview a part, and use that part to draw conclusions about the whole.

Think of it this way: imagine you want to know the opinion of residents of a city of 500 thousand inhabitants about a new urban mobility project. Talking to all 500 thousand would be practically impossible, expensive, and time-consuming. So you select, for example, 1.000 people who represent this population well, with different ages, neighborhoods, and social profiles. These 1.000 residents are your survey sample.

The logic behind this is statistical: if the sample is well chosen, what it shows tends to reflect, with a known margin of error, what would happen if you asked everyone. This is the magic (and science) behind electoral surveys, satisfaction surveys, academic studies, and even product tests.

It is worth highlighting the difference between two terms that are sometimes confused:

• Population (or universe): is the total set of people, companies, products, or events you want to study.
• Sample: is the subset of that population that you will actually survey.

The more clearly defined this distinction is at the beginning of your project, the easier it will be to design the rest of the methodology.

Why is the sample important in a survey?

The sample is, without exaggeration, one of the most important decisions in any survey. It is the foundation upon which all results will rest. If the sample is poorly chosen, not even the most sophisticated statistical analysis in the world will save the conclusions of your study. It's the famous “garbage in, garbage out.”

Here are the main reasons why the sample deserves so much attention:

1. Time and resource savings. Surveying an entire population is often financially and operationally unfeasible. A well-sized sample delivers reliable results at a much lower cost.
2. Representativeness. A good sample reflects the characteristics of the population: age, gender, income, location, purchasing behavior, among other variables relevant to your objective. This allows for safe generalization of results.
3. Statistical validity. The statistical methods used to calculate margins of error, confidence intervals, and significance depend on the sample having been selected appropriately. Without this, these numbers lose their meaning.
4. Safer decision-making. Companies use market research to decide on product launches, price adjustments, and communication strategies. If the sample is not reliable, the decision based on it will not be either.
5. Study credibility. In academic, scientific, or journalistic research, how the sample was defined is often one of the first points analyzed by those evaluating the work. A poorly done sample compromises the credibility of everything that follows.

In summary: the sample is the bridge between what you ask and what you can safely state about reality. The more solid this bridge, the more confidence people will have in your results.

What are the types of sampling?

There are two main families of sampling: probabilistic and non-probabilistic. The central difference between them lies in how participants are selected and, especially, in the possibility (or not) of calculating the probability of each person in the population being part of the sample.

Probabilistic sampling

In probabilistic sampling, every element of the population has a known and non-zero chance of being selected. This allows for calculating margins of error and generalizing results with statistical rigor. It is the most suitable type when the objective is to produce conclusions representative of the entire population.

Simple random sampling

It is the most intuitive method: all elements of the population have exactly the same chance of being chosen, as in a lottery. If you have a complete list of 10 thousand customers and randomly select 300 of them (for example, using a random number generator), you are doing simple random sampling.

• Advantage: simplicity and absence of selection bias.
• Limitation: requires a complete list of the population, which is not always available.

Stratified sampling

In stratified sampling, the population is divided into subgroups (strata) with common characteristics, such as age group, region, or education level, and then random sampling is performed within each stratum, respecting the proportion of each group in the total population.

For example, if 60% of your customers are women and 40% are men, your sample should maintain the same proportion. This ensures that smaller, but relevant, subgroups are not underrepresented.

• Advantage: greater precision when there are important differences between subgroups.
• Limitation: requires prior knowledge of the population's characteristics to define the strata.

Systematic sampling

In this method, you define a fixed interval (for example, every 10 people) and select elements from the population based on this interval, starting from a random starting point. If you have a list of 1.000 customers and want a sample of 100, you can select one customer every 10, starting from a number drawn between 1 and 10.

• Advantage: simpler to apply than simple random sampling when there is an organized list.
• Limitation: can generate bias if there is some hidden pattern in the list that coincides with the chosen interval.

Cluster sampling

Instead of selecting individuals directly, you divide the population into natural groups (clusters), such as neighborhoods, schools, or branches, randomly select some of these groups, and interview all or part of the people within the selected groups.

For example, instead of randomly selecting people throughout an entire city, you randomly select 10 neighborhoods and interview all residents in those neighborhoods.

• Advantage: reduces logistical costs, especially in face-to-face and geographically dispersed surveys.
• Limitation: tends to have a larger margin of error than other probabilistic methods, because the clusters may not be as homogeneous as the total population.

Non-probabilistic sampling

In non-probabilistic sampling, not every element of the population has a known chance of being selected, and selection usually depends on practical criteria or the researcher's judgment. It is faster and cheaper, but the results cannot be generalized with the same statistical rigor as probabilistic sampling.

Convenience sampling

The researcher selects participants who are most accessible at the moment, without probabilistic criteria. A classic example is interviewing people passing in front of a shopping mall at a certain time.

• When to use: exploratory studies, pilot tests, or when time and budget are very limited.
• Caution: high risk of bias, as the accessible group may not well represent the total population.

Purposive sampling (or judgmental sampling)

The researcher deliberately chooses participants who have specific characteristics relevant to the study's objective. It is common in qualitative research, when seeking to deeply understand the opinion of experts or a very specific group.

• When to use: qualitative research, case studies, interviews with experts.
• Caution: heavily depends on the researcher's judgment, which can introduce selection bias.

Quota sampling

Similar to stratified sampling, but without random selection: the researcher defines quotas (for example, 50 men and 50 women) and fills these quotas with whoever is available, without random selection within each group.

• When to use: quick market research, when you want to ensure basic demographic representativeness without the cost of full probabilistic sampling.
• Caution: there is still bias in how people within each quota are chosen.

Snowball sampling

Mainly used to study populations difficult to access directly, such as vulnerable groups or very specific communities. The researcher starts with a few participants and asks them to indicate other people with the desired profile, forming a “snowball” that grows with each referral.

• When to use: studies with hard-to-reach or publicly less visible populations.
• Caution: the sample tends to reflect only the social networks of the first participants, which can limit the diversity of results.

How to define the sample size?

Defining the ideal sample size is one of the most technical steps in the process, but the logic can be understood without being a statistician. The calculation depends on four main elements:

1. Population size. How many people (or items) exist in the total universe you want to study. For very large populations, this number has little impact on the final sample size beyond a certain point.
2. Confidence level. Indicates how much you can trust that the sample result reflects the reality of the population. The most common levels are 90%, 95%, and 99%. A confidence level of 95% means that if you repeated the survey multiple times, in 95% of the times the actual population result would be within the calculated margin of error.
3. Margin of error. This is the acceptable range of variation between the sample result and the actual population result. Common margins range between 3% and 5%. The smaller the desired margin of error, the larger the sample needs to be.
4. Expected variability (or heterogeneity). Refers to how much responses tend to vary within the population. When this variability is unknown, the most conservative value, which is 50%, is usually used, as it requires the largest sample size (ensuring calculation safety).

The classic formula for finite populations is:

n = (N × Z² × p × (1-p)) / (e² × (N-1) + Z² × p × (1-p))

Where:

• n = sample size
• N = population size
• Z = value corresponding to the confidence level (1,96 for 95%, for example)
• p = expected proportion (usually 0,5, when there is no previous data)
• e = margin of error (for example, 0,05 for 5%)

In practice, you don't need to do this calculation by hand. There are several free sample calculators online where you just need to enter the population size, the desired confidence level, and the margin of error to get the ideal number of participants.

A quick example: for a population of 100.000 people, with a confidence level of 95% and a margin of error of 5%, the recommended sample size is around 383 participants. If you reduce the margin of error to 3%, this number rises to about 1.056 participants, showing how the desired precision directly impacts the collection effort.

How to choose a sample for a survey?

Choosing the right sample involves answering, in the correct order, some key questions: what is the objective of the research?

Before thinking about numbers, clearly define what you want to discover. Exploratory research, which seeks to understand general perceptions, usually accepts non-probabilistic samples. Conclusive research, which seeks to generalize results, requires probabilistic sampling.

Who is the target population?

Precisely delimit who is part of the research universe. “Brazilian consumers” is too vague; “women aged 25 to 40, residents of capital cities, who bought clothes online in the last 3 months” is a much more useful cut.

Is a complete list of the population available?

If so, probabilistic methods such as simple random or systematic sampling become viable. If not, it may be necessary to resort to non-probabilistic methods or cluster sampling.

What is the available budget and deadline?

Probabilistic sampling usually requires more time and resources. If the deadline is short, quota or convenience sampling may be more realistic, provided the limitations are clear in the final report.

Is the study qualitative or quantitative?

Qualitative research, which seeks depth instead of generalization, usually uses smaller purposive samples. Quantitative research, which seeks representative numbers and percentages, depends on larger probabilistic samples.

Which variables are essential to represent?

If age, gender, income, or location are relevant factors for your objective, consider stratified sampling to ensure these groups are proportionally present.

After answering these questions, the natural path usually becomes much clearer: you already know whether you need a probabilistic or non-probabilistic sample, you have already delimited the population, and you already have an idea of the necessary size.

Survey sample example

Let's look at some practical examples of survey samples, applied to everyday situations, to make all this more concrete.

Example 1: E-commerce customer satisfaction survey

• Population: all customers who purchased on the website in the last 12 months (assuming 50.000 people).
• Sampling type: stratified, dividing customers by spending range (low, medium, high) to ensure all consumption profiles are represented.
• Sample size: with a confidence level of 95% and a margin of error of 5%, about 381 customers.
• Collection method: email with a link to an online questionnaire, sent proportionally to each spending bracket.

Example 2: Academic research on university students' reading habits

• Population: students enrolled in a university with 20.000 students.
• Sampling type: simple random, drawing names from the official enrollment list provided by the institution.
• Sample size: approximately 377 students, considering 95% confidence and 5% margin of error.
• Collection method: online form sent via institutional email.

Example 3: Market research for a new product launch

• Population: potential consumers within a specific age group and region, without a complete list available.
• Sampling type: quota sampling, defining quotas for age, gender, and income based on IBGE demographic data for the region.
• Sample size: 400 participants, distributed proportionally among the defined quotas.
• Collection method: approach at high-traffic locations, such as shopping malls and commercial centers.

Example 4: Qualitative research with industry experts

• Population: professionals with over 10 years of experience in a specific technical niche.
• Sampling type: purposive, selecting recognized names in the market based on predefined criteria (years of experience, position, publications).
• Sample size: between 8 and 15 interviewees, a typical number for qualitative studies that seek depth, not statistical generalization.
• Collection method: individual in-depth interviews, in-person or via video call.

Example 5: Study on a hard-to-reach community

• Population: homeless people in a specific city.
• Sampling type: snowball, starting with a few contacts made through NGOs and asking for referrals of other people with the same profile.
• Sample size: variable, defined according to the saturation of responses (when new interviews no longer bring new relevant information).
• Collection method: in-person interviews conducted by trained teams, often in partnership with social assistance.

These examples show how the choice of sampling type changes completely depending on the objective, population, and available resources. There is no single “right way” to do sampling: there is the most appropriate way for each specific context.

What errors should be avoided when defining the sample?

Some errors appear so frequently in poorly planned research that it is worth highlighting them specifically.

• Confusing sample size with representativeness

A large but poorly selected sample can be less reliable than a smaller but well-distributed sample. Size matters, but the selection method matters just as much, if not more.

• Ignoring selection bias

When the method of recruiting participants systematically favors a certain profile, the result is distorted. A classic example is conducting research only on social media, which automatically excludes those who do not use that platform.

• Not clearly defining the target population

Without a precise definition of who should be part of the research, it becomes impossible to assess whether the sample is adequate or not.

• Using convenience sampling when the objective requires generalization

Convenience samples are useful for exploratory studies but do not allow for stating “this represents the entire population” with statistical certainty.

• Forgetting the response rate

If you plan a sample of 400 people, but know that only 20% usually respond to surveys in your sector, you need to invite a much larger number of participants to reach the desired final sample.

• Not considering relevant subgroups

If characteristics such as age, gender, region, or income directly influence the researched topic, ignoring these subgroups in the sample composition compromises the quality of the results.

• Mixing methods without transparency

Combining probabilistic and non-probabilistic sampling can be valid in certain research designs, but this needs to be clearly explained in the report so that those interpreting the results understand the limitations involved.

• Defining the margin of error and confidence level after collecting the data

The ideal is to plan these parameters before collection, and not adjust the numbers afterward to “fit” a result that has already been obtained.

Avoiding these pitfalls requires planning, but the effort is worth it: a well-defined sample saves time, money, and, most importantly, prevents decisions made based on data that do not reflect reality.

Looking for people for your research sample?

At PainelTAP, your research finds the right people quickly, with quality and security. Request a proposal and discover how to optimize your data collection.

I want to know more

Frequently asked questions

How many people should participate in a survey?

There is no fixed number valid for all surveys. The ideal depends on the population size, the desired confidence level (usually 95%), and the acceptable margin of error (usually between 3% and 5%). For large populations, samples between 380 and 1.100 people are usually sufficient for most quantitative studies, but the exact number should be calculated on a case-by-case basis, using a sample calculator or the appropriate statistical formula.

What is a representative sample?

A representative sample is one whose characteristics proportionally reflect the characteristics of the population it intends to represent. This includes variables such as age, gender, location, income, and other factors relevant to the study's objective. When a sample is representative, the results obtained can be generalized to the entire population with a known and acceptable margin of error.

What is the difference between census and sampling?

A census consists of collecting data from absolutely all elements of a population, without exception, as happens in the IBGE Demographic Census. Sampling, on the other hand, consists of selecting only a part of the population to represent the whole. A census tends to be more accurate, but much more expensive, time-consuming, and complex to execute. Sampling, when well-planned, offers reliable results with much lower cost and time, being the most common choice in the vast majority of market, academic, and opinion research.

In conclusion…

Properly defining a research sample is not just any technical detail: it is the foundation that supports the credibility of any study, whether academic, market, or public opinion. Understanding the available sampling types, knowing how to calculate the ideal size, and being aware of the most common errors allows any research, of any size, to deliver results that can truly be trusted.

If you are about to plan your own research, it is worth reviewing each research sample example presented here and identifying which method best fits your context, your budget, and your final objective. With this initial care, the rest of the process—data collection, analysis, and interpretation—tends to flow much more smoothly and reliably.