Stratified Sampling: Definition, Formula, Examples, Types

Stratified sampling is a probability sampling technique that divides a population into distinct subgroups called strata, and draws a random sample from each one. It is used when a population contains meaningfully different segments whose characteristics are relevant to the research objective.

The advantage over simple random sampling is precision. By ensuring every subgroup is represented in proportion to its presence in the population, stratified sampling produces results that are more accurate, more reliable, and more useful for comparative analysis across segments.

What we’ll cover in this article:

1. Stratified Sampling: Definition
2. How does stratified sampling work
3. Formula for stratified sampling
4. Stratified sampling Vs cluster sampling
5. How to perform stratified sampling
6. Stratified sampling examples
7. Types of stratified sampling
8. The pros and cons of stratified sampling

Stratified Sampling: Definition

Stratified sampling is a probability sampling technique where the total population is divided into distinct subgroups called strata. Each stratum is made up of individuals who share similar characteristics — age, gender, income, education level, or any other attribute relevant to the research objective. Once the strata are defined, a random sample is drawn from each one.

The difference between stratified sampling and simple random sampling comes down to control. 

In simple random sampling, every individual in the population has an equal chance of being selected — which sounds fair, but can result in certain subgroups being over or under-represented purely by chance. Stratified sampling removes that uncertainty. By dividing the population first and sampling within each group separately, it guarantees that every subgroup appears in the final sample in the right proportion. 

This matters most when the subgroups you care about are meaningfully different from each other — because a sample that misrepresents them produces conclusions that do not reflect reality.

A quick example

A researcher wants to understand the likelihood of marriage among adults with similar education levels. The population is divided into two strata based on gender — male and female. A random sample is then drawn from each stratum in proportion to their representation in the total population. The results from both subgroups are compared to draw conclusions.

Gender is the stratum. (stratum: singular for strata)

The two subgroups are mutually exclusive. And because the samples are drawn randomly within each stratum, the findings are representative of each group rather than skewed toward whichever gender happens to be more prevalent in a general random sample.

How Does Stratified Sampling Work?

Stratified sampling works by dividing a population into mutually exclusive subgroups before any sampling takes place. Every individual in the population belongs to exactly one stratum (there is no overlap). Once the strata are defined, a random sample is drawn from each one independently, and the results are combined to form the final study sample.

The process follows three stages.

Stage 1: Define the strata

The researcher identifies the characteristics that divide the population into distinct groups. These characteristics should be directly relevant to the research objective. 

For example, gender, age bracket, income range, education level, and geographic region. The quality of the strata determines the quality of the final sample. Poorly defined strata produce a sample that is no more reliable than a general random one.

Stage 2: Draw a sample from each stratum

Once the strata are defined, a random sample is drawn from each one independently. The size of each sample is typically proportional to the stratum's share of the total population. A stratum that makes up 40% of the population contributes 40% of the total sample. This proportionality is what makes the final sample representative of the real-world population — not just statistically balanced on paper.

Stage 3: Combine and analyze

The samples from all strata are combined to form the final dataset. From here, researchers can analyze the population as a whole or compare how individual strata respond differently to the same questions. That second level of analysis — comparing across subgroups — is often where the most useful insights emerge. It is the reason stratified sampling is preferred over simple random sampling when the research objective involves understanding differences between segments.

Stratified Sampling: Formula

The selection of the target market, strata, and the total sample size in stratified sampling does not require a formula. That’s at the researcher’s discretion.

However, we use this formula to find the sample size of each subgroup involving one or more strata:

Stratified sub-group sample size = (Total Sample Size / Entire Population) * Population of Subgroups (Stratum Size) 

For example, you’re interviewing a school to understand the type of food that the students like. This school has both boys and girls, and you want to take their thoughts into account with a sample size of 100 students.

Here are the numbers:

• Total Students: 2,000
• No. of boys: 800
• No. of girls: 1,200

So, using the stratified sampling formula;

• Total girls in the final sample: (1,200 / 2,000) * 100 = 60
• Total boys in the final sample: (800 / 2,000) * 100 = 40

Hence, using this formula, you get the size of each sub-group in proportion to the size of each sub-group (girls and boys) where the stratum is gender.

Stratified Sampling Vs Cluster Sampling

Stratified sampling and cluster sampling are both methods of dividing a population into subgroups before sampling. The similarity ends there. How those subgroups are formed, and how samples are drawn from them, are fundamentally different — and choosing the wrong method for a research objective produces unreliable results.

How the groups are formed

In stratified sampling, the researcher defines the subgroups based on shared characteristics — age, gender, income, or any attribute relevant to the research. Every individual in the population is assigned to a stratum based on those characteristics, and the strata are designed to be as internally similar as possible.
In cluster sampling, the population is divided into naturally occurring groups — geographic regions, schools, hospitals, or organizational units. These clusters are not defined by shared individual characteristics. They are defined by proximity or membership. A cluster might contain individuals with very different ages, incomes, and behaviors — and that heterogeneity is intentional.

How samples are drawn

In stratified sampling, a random sample is drawn from every stratum. No stratum is left out. This is what guarantees representation across all segments of the population.

In cluster sampling, only some clusters are selected — typically at random — and every individual within the selected clusters is surveyed. Clusters that are not selected contribute no data to the study at all.

When to use each

Stratified sampling is the better choice when the research objective requires accurate representation of specific subgroups, or when comparing how different segments respond to the same questions. It produces more precise results but requires detailed information about the population's composition upfront.

Cluster sampling is more practical when the population is geographically dispersed and surveying every individual would be prohibitively expensive or logistically difficult. It trades some precision for cost and operational efficiency.

How to Perform Stratified Sampling

Following are the crucial steps in conducting stratified random sampling.

• 1: Know All Attributes
• 2: Decide Your Groups Of Interest
• 3: Finalize The Sample Size
• 4: Create Random Sample
• 5: Survey!

Step 1: Know All Attributes

With stratified sampling, your team should be clear about the attributes (strata) they’ll be targeting to compare. So, they should know about relevant subgroups within the market that may create different strata due to behaviors and characteristics.

Suppose the target market is school students and you wish to understand what books they like to read. Now, there can be multiple strata here, like:

• Gender
• Age
• Interests
• Academic record
• Type of teachers
• Parent’s qualification

These and more subgroups will form in this case. We included the ‘type of teachers’, and ‘parents’ qualifications’, as these influence students’ choices massively. But the bottom line is that your team needs to know all attributes in a target market to conduct market research using stratified sampling.

Step 2: Decide Your Groups Of Interest

So, continuing with the same example, if you wish to understand the type of books girls prefer after considering the influence of their parent’s educational qualification, the strata will be ‘girls’ and ‘parent’s qualification’ for study out of the total population. Similarly, ‘boys’ and ‘parents’ qualifications’ can form another sample.

We can create more subgroups as per the requirements of your market research. Like only girls over 15 years or parents with at least a master’s degree. This way, the final sample that comes out is much more in-line with your research requirements. Hence the results are more accurate and reliable.

Step 3: Finalize The Sample Size

After finalizing different samples for different strata and sub-group combinations, you and your team decide the size of these samples. We find this by looking at your research objectives.

The sample sizes may or may not be the same in numbers, as it depends on the total population (target market). But the ratio of selection from each stratum will be proportional.

Step 4: Create Random Sample

You chose different strata, identified further sub-groups, and finalized the sample size. All that remains is to calculate the sample size using the stratified sampling formula and form all samples.

Remember, all samples will be mutually exclusive if you apply the entire process, including the formula, correctly.

Step 5: Survey!

After creating all stratified samples, the last step is to go ahead and conduct surveys. SurveySparrow’s market research survey software lets you conduct highly focused surveys and analyze the results for all samples in real-time.

With the Audience Panel, your team needn’t worry about collecting and then analyzing the survey results. They just need to select the right question types, customize them, hit ‘send,’ and then wait for the results.

They can also analyze responses at both the overall and the individual levels. 

Would you like to see how SurveySparrow’s dashboard looks for a market research survey?

Get Your Sample Market Research Report

This interactive report will give you an idea of how SurveySparrow presents your data visually and helps you get actionable insights

Send Report

Stratified Sampling: Examples

Now that you know the steps in the stratified sampling process, it’s time to let the cat out of the bag and give examples for this sampling technique. On we go, then.

1. For Extending Offerings

For a local lawn care company hoping to expand offerings, a stratified sample survey can give a credible answer on the specific products to offer.

If the target audience is existing customers, they can divide relevant subgroups based on age range, gender, annual income, and if they already own products from the company.

The resulting data and insights will give a better idea of the specific groups to target in this case.

2. For A Campaign’s Success

For a campaign’s success, your team needs to nail down the messages that’ll resonate most effectively with a very specific target population.

Yet, you also want to understand how to connect with people from other subgroups, such as those with household incomes under the national average, those who’re living a sedentary lifestyle, etc. Market research using stratified sampling is used extensively in these cases, and rightly so.

For Getting New Customers

For appealing to new customers for your brand, conducting market research using stratified sampling works great.

Let’s say you have a loyal following of customers above the age of 40, but you wish to attract younger customers for more sustained growth. So, your overall target population can be people in the age group of 25-40 who have purchased or enquired about your products at least once in the past year.

Relevant strata and sub-groups will give a final sample population that’s ready for market research to find out how you can attract a younger audience.

For Timely Estimations

We’ll give a real-life example here. A research paper published in medRxiv discusses the suitability of using the stratified random sampling technique for estimating COVID-19 prevalence in the U.S. state of Maryland.

In this survey, the population of Maryland was stratified or divided based on counties. Then, individuals were selected from each county representing their stratum.

As per the study, the stratified sampling technique for testing COVID-19 prevalence is acceptable. But the sample arrived through stratification must be adjusted for misclassification error to avoid under-or overestimation of COVID cases. See how crucial a role this sampling technique played here?

For Gauging Employee Satisfaction

Yes, stratified sampling is used here, too. The target population is your employees, and the stratum selected is their satisfaction levels, based on the employee feedback surveys.

SurveySparrow’s employee 360 degree surveys are also used for it, as it gets overall information on an employee’s performance.

Based on the satisfaction levels gained here, unhappy employees are put in the final sample to conduct surveys that can reveal the reason for their dissatisfaction and ways to change that.

Stratified Sampling: Types

The main aspect of stratified sampling is that every stratum is different from the others. When we form subgroups from these strata, they should all be mutually exclusive. To achieve that, your team needs to rely on two stratified sampling methods or types. Let’s discuss both:

Proportionate Stratified Sampling

In proportionate stratified sampling, we select the stratum size for the final sample based on their original distribution in the population of interest. Therefore, strata that are less represented in the total population will also find fewer occurrences in the final sample.

With this approach, your team can prepare a final sample that majorly represents the dominant stratum, giving them an in-depth understanding. Out of the two stratified sampling types, this is predominant in highly-focused market research, where there are clear instructions on the type of strata required in the sample group.

When using stratified random sampling to select 500 graduates by age group, the proportional formula is:
(Sample Size ÷ Population Size) × Stratum Size.
This ensures every age group is properly represented, delivering balanced and trustworthy data.

Disproportionate Stratified Sampling

In disproportionate stratified sampling, every unit in a particular stratum from the total population has a similar chance of getting into the final sample population. So, no predominant subgroup has a higher chance of getting selected. 

This stratified sampling type is often used to study underrepresented subgroups in the target market.

Basically, the entire sample selection process rests in the hands of the researcher or survey taker, or your team, as they can select as many people as they like from a particular subgroup based on the research requirements. So, taking the same example as above, the researcher could select 1,500 males and 500 females or vice-versa when using disproportional stratified sampling for market research.

Pros & Cons of Stratified Sampling

Here we are, then. We know stratified sampling with its definition, formula, examples, and types. The one thing that remains now is to talk about the advantages and disadvantages of this sampling technique. Time to do that.

Pros

• More representative sample. By dividing the population into subgroups before sampling, stratified sampling ensures every meaningful segment is included in the final sample. A general random sample might miss smaller but important subgroups entirely — stratified sampling prevents that.
• Reduced sampling error. Because each stratum is sampled independently, the final sample reflects the actual composition of the population more accurately than simple random sampling. This reduces the margin of error in the findings, particularly when subgroups behave differently from one another.
• Enables comparative analysis. Stratified sampling makes it straightforward to compare how different segments respond to the same questions. This is often where the most actionable insights emerge — understanding not just what the overall population thinks, but how specific groups differ from each other.
• More efficient use of resources. Studying an entire population is rarely practical. Stratified sampling produces reliable results from a smaller, well-constructed sample — saving time and research budget without sacrificing the accuracy needed to make confident decisions.
• Reduces bias in analysis. Because every stratum receives proportional representation, no single group dominates the findings. This produces a more balanced analysis and reduces the risk of conclusions that reflect one segment's experience more than others.

Cons

• Requires detailed population data upfront. To divide a population into strata, the researcher needs accurate information about its composition. If that data is incomplete, outdated, or unavailable, the strata cannot be defined correctly — which undermines the reliability of the entire sample.
• Strata selection directly affects results. The choice of which characteristics to use as strata is consequential. Include the wrong attributes and the subgroups formed will not reflect the meaningful differences in the population. There is no objective formula for this decision — it requires careful judgment and a clear understanding of the research objective.
• Not suitable for small populations. When the total population is small — generally fewer than 100 individuals — sampling is unnecessary. The entire population can be studied directly. Stratified sampling adds complexity without adding value in these situations.
• Researcher bias can influence the sample. Decisions about which strata to include, which subgroups to prioritize, and how many individuals to select from each stratum are all made by the researcher. Those decisions introduce the possibility of bias, particularly when underrepresented subgroups receive less attention than their importance to the research warrants.
• More complex to design and execute. Stratified sampling requires more planning, more detailed population data, and more careful sample construction than simple random sampling. For research teams without experience in the method, this complexity increases the risk of errors that affect the validity of the results.

Wrapping Up

Whether you go with proportionate or disproportionate type stratified sampling, the most crucial part is creating internally homogenous sub-groups. These are mutually exclusive with other sample populations that were created using different stratum and sub-groups.

This way, you can have a fair share of minority groups in the final sample for holistic market research.

Also, avoid biases and giving too much attention to one of the sub-groups in your stratified samples, unless the research is specifically asking for it. This skews your sample and distorts the expected results.

And once you have a well-rounded final sample, you can use SurveySparrow to deploy market research surveys and get the information you need. So, deploy stratified sampling in your next market research. It’ll be worth the effort, and we’ll be waiting to hear all about it.