This article describes the most common probability sampling methods used for the primary market research purposes.
Sampling exists and is necessary because surveying the entire population (all the potential customers of a particular market) is neither practical nor possible due to lack of resources. Besides, this would be too expensive and extremely time-consuming.
What should the sample look like?
Businesses select a sample for a particular market – a small group of people that is truly representative of the whole population. The population is all the potential customers of a product, or the overall target market.
A sample size of approximately 1,000 respondents is sufficient as it will very likely produce fairly accurate results.
When designing a sample, the business organization needs to consider the sample frame, a list of all people in the target population from which the sample is drawn, and the sample size, the number of people to include in the research.
Sampling results must be statistically accurate as determined by the confidence level of the results of .95 (95%) or higher.
The two broad categories of sampling include:
1. PROBABILITY SAMPLING METHODS:
a. Simple random sampling
b. Systematic sampling
c. Quota sampling
d. Stratified sampling
e. Cluster sampling
f. Multistage sampling
2. NON-PROBABILITY SAMPLING METHODS:
a. Convenience sampling
b. Snowballing sampling
c. Judgmental sampling
When choosing between different sampling methods, market research specialists need to determine who and what needs to be asked in the research. Then, when a chosen sampling method allows to generate statistically valid representative responses to the research question in a cost-effective way, this confirms that the choice was right.
1. PROBABILITY SAMPLING METHODS
The probability of each respondent’s inclusion in the sample can be calculated. Respondents have a random chance to be chosen.
All of the following probability samples are likely to lead to more accurate results – which are more representative of the whole population – than non-probability sampling techniques. Also, the chances of any errors occurring can be estimated as well.
The most common probability sampling methods include the following six methods:
a. Simple random sampling
Simple random sampling gives all people (members of the target population) an equal chance of being selected and included in the sample.
How to conduct simple random sampling?
These days the computer software is picking numbers representing each person out of a hat – randomly. Hence, a list of all of the people in the target population is needed. The population is organized in a database with the unique number assigned to each individual person. Then, a list of random numbers is generated by the computer. The numbers on the random number list are selected and the people who had these numbers allocated to them will form the sample up to the required number of respondents.
TIP: You can use the random number function (RAND) in Microsoft Excel to generate random numbers.
When to use simple random sampling?
Simple random sampling is used when the population is homogenous. It means that all members of the target market have exactly the same, or very similar, characteristics. For example, all parents of children at the school, university students enrolled at the particular university, all passengers flying on the economy-class.
Advantages of simple random sampling
Simple random sampling is quite easy to get a totally unbiased sample. When everyone has an equal chance of being selected, any bias or unrepresentative samples will be prevented.
Disadvantages of simple random sampling
Simple random sampling assumes that everybody in the population is the same, or very similar, therefore, due to the randomness of choice, it might include people in the sample who are not really a part of the target market. Also, samples need to be large enough to generate representative results, hence collecting a list of all of the people in the target population will be costly and time consuming.
COMMON MISTAKE: Do not mix up random sampling and convenience sampling. Selecting the first ten people who enter a shop is not a random sample as every member of the population (potential customers) the firm is interested in does not have an equal chance of being selected. Only those entering the shop stand a chance of being chosen. Indeed, those buying from this shop may have similar characteristics not shared by others in the sampling frame. This mistake is often made as people often say ‘I chose the first ten people at random’. That is incorrect!
b. Systematic sampling
Systematic sampling, which is no longer random, chooses people from a larger population using a fixed periodic interval such as every 5th person, 10th person, 20th person, etc.
How to conduct systematic sampling?
While the sampling frame is composed randomly, respondents for the sample are chosen at regular interval. Every nth item from the target population is taken until the desired size of sample is reached. When the periodic interval is every 5th person, the random starting point should be between 1 and 5 to ensure that in practice systematic sampling has features of randomness. Each of the first five people has an equal chance of being chosen though. If you want to sample 10 employees from a workforce of 200, then 200 needs to be divided by 10. So, every 20th worker is chosen after first choosing a random starting point between 1 and 20.
When to use systematic sampling?
Systematic sampling can be chosen when a retail business wants to study purchasing patterns of its customers visiting the outlet. Or, the supermarket staff can ask every nth person entering inside until reaching the required sample size.
Advantages of systematic sampling
Systematic sampling is easier to conduct than simple random sampling and it spreads the sample more evenly over the population. Also, it randomly selects the starting point hence can provide somewhat unbiased sample.
Disadvantages of systematic sampling
Some numbers that are selected to be chosen may not exist while other respondents from the end of the list of all respondents may never be chosen. Also, there might be some hidden regular patterns in the population that will skew the system.
c. Quota sampling
Quota sampling divides the entire population into different groups (quotes) who share common characteristics such as the same gender, age, occupation, income, education level, social status, interests, values, etc. Then, a certain number of people from each segment is selected.
How to conduct quota sampling?
Quota sampling involves segmentation. The total population is first split into different segments that share similar characteristics. Then, a quota (certain number of people) from each market segment is selected. The researcher makes the decision which characteristic to use to group the sample, and how many people will be chosen. For example, 50 men between the age of 20 and 40 as well as 50 men between the age of 41 and 60 will be selected to make up the sample of 100 respondents. Finally, quotas will be filled from each specified sub-groups of the population. The selection of the sample is done in a non-random way.
FACT: A number of respondents in each quota is not proportional to the population.
When to use quota sampling?
Quota sampling is used when the population is heterogenous, so it needs to be subdivided into different segments that share very similar characteristics. The firm wants to get feedback from different groups of people, but it will not represent the views of the entire population.
For example, let’s say that the company may want to interview different types of workers about pay levels and working conditions. There are 800 workers currently hired at that business. The marketing manager decided to divide staff into the following four groups to come up with the sample of 100 workers:
Group 1: Men working full-time 25 (25% of the sample)
Group 2: Men working part-time 25 (25% of the sample)
Group 3: Women working full-time 25 (25% of the sample)
Group 4: Women working part-time 25 (25% of the sample)
The allocations of workers into different segments are not done proportionally, therefore are not representative of the population. Interviewees are not selected according to the different proportions that certain worker groups make up of the whole workforce. The real proportion of men working full-time at that business is ignored.
Advantages of quota sampling
Quota sampling is extremely useful. It allows to obtain a fairly representative sample quickly and easily, hence cheaply. By including respondents from different market segments, the research findings will be more reliable than simply picking anyone to ask on the street in unstructured way.
Disadvantages of quota sampling
Samples generated using quota sampling are not always representative of the population as a whole. The selection of respondents into the sample is non-random, therefore not everyone gets an equal chance of being selected. The number of people chosen in each segment is just guesswork not being based on any statistical information about the population. Also, the researcher might be biased in selecting people in each quota preferring to ask only tall and visually-attractive people, for example.
TIP: Quota sampling is very similar to stratified sampling. However, in quota sampling, respondents are selected in a non-random way. While in stratified sampling, random sampling is used to select an appropriate number of respondents for each stratum.
d. Stratified sampling
Stratified sampling divides the entire population into different groups (strata) who share specific characteristics such as the same gender, age, occupation, income, education level, social status, interests, values, etc. Then, a certain number of people from each segment is selected according to the same percentage in total population. If the population includes 40% of men working full-time, so would the sample.
How to conduct stratified sampling?
Stratified sampling involves segmentation, proportional allocation and then simple random sampling within each stratum. Strata means layers of the population.
STEP 1: The precise knowledge about how population is divided up is necessary. The sampling frame is split into different non-overlapping segments that share similar characteristics.
STEP 2: Then, a strata (certain number of people) from each market segment is selected that is proportional to the population. The researcher does not make the decision how many people will be chosen for each strata. In general, the size of the sample in each stratum is taken in proportion to the size of the stratum in the entire population. Or, according to the different proportions that certain customer groups make up of the whole target market. This is called proportional allocation.
STEP 3: Finally, all strata will be filled from each specified sub-groups of the population. The selection of the sample is done in a random way, hence sampling is called stratified random sampling.
How to choose a sample size for each stratum?
For example, let’s say that the company may want to interview different types of workers about pay levels and working conditions. There are 180 workers currently hired at that business. Therefore, the entire population will be divided into the following four groups:
Group 1: Men working full-time 90 (50% of the population)
Group 2: Men working part-time 18 (10% of the population)
Group 3: Women working full-time 9 (5% of the population)
Group 4: Women working part-time 63 (35% of the population)
The marketing manager decided to take the sample of 40 workers. The allocations of workers into different segments must be done proportionally, therefore are representative of the population. This tells us that in the sample of sample 40, the proportions of different groups of workers will be as follows:
Group 1: Men working full-time 20 (50% of the sample)
Group 2: Men working part-time 4 (10% of the sample)
Group 3: Women working full-time 2 (5% of the sample)
Group 4: Women working part-time 14 (35% of the sample)
Interviewees are selected according to the different proportions that certain worker groups make up of the whole workforce.
When to use stratified sampling?
Stratified sampling is used when the population is heterogenous, so it needs to be subdivided into different segments that share very similar characteristics. The firm wants to get feedback from different groups of people which will represent the views of the entire population.
For example, let’s say that your company wants to survey 100 current customers about soft drink preferences during holidays. Instead of asking the first 100 customers who pick up the phone, it would be more accurate to split up all customers into certain strata such as the age. So, if the total number of customers is 1,000 of whom 50 are teenagers, an accurate sample of 100 would contain 5 teenagers. Then, repeat the process with all age groups until the total required sample of 100 was reached.
Also, a company which plans to market its product to only one market segment could use stratified sampling. For example, designer female clothes for students aimed at 16−24-year-olds. Then, only young women from this stratum of the population will be included in the market research sample.
Advantages of stratified sampling
Stratified sampling is very representative as it ensures better coverage of the population than simple random sampling. And, more precise proportionality of different strata of the total population in the sample than quota sampling. Also, findings will be more relevant with less sampling errors as the selection of respondents in each stratum is done using simple random sampling. In addition, the results from each stratum can be analyzed separately providing great insights into a particular market segment. Different consumer profiles can be established thanks to stratification as members of the same stratum are as similar as possible. This can be further use in Differentiated Marketing strategy to devise various marketing mixes for different market segments.
Disadvantages of stratified sampling
Stratified sampling is not suitable for homogenous populations. It can be very complicated. And expensive to obtain data about different proportions in the population hence making identifying appropriate strata difficult. Also, researcher may avoid using simple random sampling in their selection of people in each stratum preferring to ask only those people who are willing to participate. This will make stratified sampling more biased, hence less probability-based. Market research specialists may need to be specifically trained to deal with a particular stratum which will increase costs of market research. Finally, it will be more complex to organize and analyze results.
RECOMMENDATION: A larger sample should be used for those strata with greater variability in order to obtain more accurate results.
e. Cluster sampling
Cluster sampling divides the entire population geographically into different clusters (areas such as country, province or state, city) in order to choose respondents within each cluster, or from multiple areas.
How to conduct cluster sampling?
In cluster sampling, the whole population is first divided into different geographical areas. The choosing of clusters is often done using simple random sampling giving each cluster an equal chance of being chosen. Then, ideally, respondents within each cluster unit will be selected at random too. However, when the target population is too dispersed and the business does not have enough budget to visit each location, the units sampled, which are chosen in clusters, can be close to each other (e.g. houses along the same street). Or, a sample can even be taken from just one cluster (e.g. only from one city in the country instead of five different cities).
TIP: All the clusters chosen should be dissimilar with one another, so the sample is as representative as possible.
When to use cluster sampling?
Cluster sampling is often chosen for conducting opinion polls on the entire society or before governmental elections. Also, when a multinational company wants to research global attitudes towards its product, it will concentrate on just a few areas for its research instead wasting time and money on traveling around the whole world. The opinions from randomly interviewed and/or surveyed people within each of the selected clusters will be used to represent the views of the entire market. Additionally, useful for researching in a particular industry where taking a sample from just one company in one region may be representative for the whole industry
Advantages of cluster sampling
Cluster sampling makes it easier and cheaper to sample homogenous population that is widely dispersed over different geographical areas. It is not necessary to travel to every single place to sample people as characteristics of customers are the same, or very similar, everywhere. This will help to reduce costs.
Disadvantages of cluster sampling
Cluster sampling may turn out to be really biased as using just a few locations. And assuming that they represent the entire population is very likely erogenous. And when more clusters are used, this will clearly increase the research costs and prolong the time spent on analyzing and presenting data.
RECOMMENDATION: Increasing the number of clusters that are far away from each other in the sample can help to reduce bias.
f. Multistage sampling
Multistage sampling divides the entire population geographically into different clusters (areas such as country, province or state, city) just like cluster sampling does. Then, the areas, which are often chosen randomly, are subdivided further into more specific regions to choose respondents for the sample.
How to conduct multistage sampling?
After the population is divided geographically into different areas (clusters), then, a sample is drawn from a specific another area within that area such as country – province – city – district – county – street – block – house/apartment. The process can be repeated until individual households or companies are identified. Or, any other unit of interest.
When to use multistage sampling?
Multistage sampling is often used for political elections, governmental surveys on the society such as The UK Family Expenditure Survey (FES). In the first case, specific parliamentary wards are used while in the second case, the primary sampling unit is a postal code. Multistage sampling can also be used when the business is trying to subdivide the target market into rural, suburban and the city center areas.
Advantages of multistage sampling
Multistage sampling is a cheaper way than carrying out a total population census or simple random sampling. As the country is divided into a number of regions, the final sample is only concentrated on relatively few geographical areas.
Disadvantages of multistage sampling
The flexibility of the decision process in subdividing the country as well as in choosing the final areas may increase subjectivity – arbitrariness. Also, multistage sampling cuts out large portions of the population from the study, the findings can never be 100% representative of the population, thereby may not provide any useful information. In addition, shortlisted respondents for the study may simply not be available at home when the research is being date causing data to get lost.
In conclusion, sample is a small group of people chosen either in the probability way or non-probability way aiming to represent the entire population – the target market. Researching everyone in the total population would be too costly and time-consuming. Hence, market researchers will be testing on a sample to understand how the population will react to a product being marketed.
Each probability sampling method has its advantages and disadvantages, and each one will be suitable for a different business situation. The choice of the best sampling method will mainly depend on the purpose of the market research the population and financial resources of the business organization.
While disadvantages of sampling include issues with data reliability and bias, it can save time as fewer resources are required and aid marketing decision-making. Cost-effectiveness is always important in all market research decisions.
