A Primer on Research Sampling Methods in Selecting Your Study Subjects
When it comes to considering research sampling methods you often have to start at the beginning and consider what questions you are trying to so answer:
Will my product or service sell? How will my employees respond to a compensation change? What fundraising effort will be most accepted? These and many other similar questions typically drive an organization’s desire to conduct market research. If you are planning to embark on such an effort, one of the first questions you’ll need to answer is how you will collect your sample.
Two primary methods employed by market research experts are probability and non-probability sampling. Within each of these categories are subgroups that more clearly define how to obtain your sample, each carrying its own advantages and disadvantages to the researcher seeking the data.
In probability sampling, every individual in the entire population being considered has an equal chance of being selected for the survey, interview, or questionnaire. The selection process is completely random, and therefore, the sample is likely to be closely representative of the whole population. Probability sampling is best used in market research requiring quantitative results. Typically, these results of probability sampling involve a great deal of number crunching and statistical charts and graphs.
Three types of probability sampling are typically used in market research: cluster, stratified, and systematic.
Cluster sampling involves two stages. In stage one, the market researcher selects a certain number of groups or clusters of people to question or interview. In stage two, a random sample within each cluster is selected for the actual study.
Cluster sampling works best in situations where random sampling of an entire population would be too expensive, impossible, or extremely complicated. This method is a less expensive and faster way to collect market research information. However, since it’s not completely random, you are more likely to generate sampling errors.
One example of a good use for this research sampling method may involve collecting customer preferences for a large, national hotel chain. It would be difficult, expensive, and time-consuming to collect information about every customer visiting every location of a hotel chain. However, you can select a dozen locations around the country in stage one of a cluster sample, and then randomly select guests at each of those 12 locations over the course of a month for your research.
Perhaps you are collecting insights about new hotel services you’re thinking about adding. Customer preferences shared through such a cluster sample would probably be reasonably representative and usable for making such decisions. Precision is not extremely important in this case, and therefore, the cost and time savings would outweigh the need for conducting a completely randomized survey.
Stratified sampling is a method where the overall population is divided into mutually exclusive groups before a random sample is selected. You might want to sub-divide your group by gender, race, income levels, or age. Each person can only belong to one strata or group.
Businesses or organizations looking for a high level of precision or the ability to analyze information within the smaller subgroups in addition to the overall population may want to invest in stratified sampling. Since a representative group will be selected from each strata, the actual sample can be smaller, which will save time and money.
Depending on your population and research goals, you’ll want to decide if you will use a proportionate or disproportionate stratification. Proportionate stratification can increase your precision because the actual sample of people will be proportionate to your entire population, which may not be the case in a completely randomized sample. Disproportionate stratification can help market researchers when there are significant variances among the strata. You may be able to gain precision for a particular survey measure; however, this precision may not carry across other components of the research.
Systematic sampling is an easy version of probability sampling because researchers select every nth individual on a population list. As long as the population list does not contain any pre-organization, the resulting sampling should be representative.
This method of sampling is simple, fast, and effective in most market research situations. All that is necessary is a list of the population, a starting point, and an interval. For example, if you want to gather data from a trade association with 10,000 members, you can select every 100th person on a membership roster to create a survey group of 100.
One example of a potential problem with systematic sampling would be a list that is organized before the sample is selected. For instance, if you are questioning coaches and players of an adult sporting league about tournament locations, and the list is made up of team sub-lists that always place two coaches followed by 20 team members, you run the risk of either soliciting feedback from all coaches or no coaches depending on your interval selection.
Non-probability sampling is, obviously, the opposite of probability sampling. The selection process is not random, and therefore, subject to research bias and sampling errors.
The results of non-probability sampling are often helpful before or after a market research project involving probability sampling. For instance, the ideas generated can be used to create a quantitative survey for a randomized sample of your population, or a non-probability sample can help flesh out and clarify topics that come out of a randomized survey.
Three types of non-probability sampling are typically used in market research: convenience, quota, or judgmental sampling.
Convenience sampling is a quick-and-easy way to select your research subjects because they are the ones most convenient to your particular research project. This factor means that it’s faster, easier, and cheaper to conduct your research. The major disadvantage is that (depending on the type of convenience sampling you are using) you can introduce significant bias or sampling errors using this method.
One common example of convenience sampling is in a university setting where graduate students use volunteer undergraduate students as subjects for experiments. In other cases, a researcher may just select the people who happen to shop at a particular store on one day, mall shoppers, or the first dozen clients on a business’ customer list.
There are hybrids for convenience samples in online research where you draw random samples from a universe of participants with certain characteristics using behavioral data and other targeting methods. It’s a targeted convenience sample but still random. The question becomes for low incidence categories; is it really better to screen through 10,000 people to get 100 people who qualify for a 1 percent incidence study? Most likely it’s cost and time prohibitive, if you consider all the pros and cons.
Quota sampling is a non-probability sampling method that can be valuable in particular market research efforts. Using this technique, a researcher will sub-divide a population to study a particular group within that population. The best use of quota sampling is to research a particular trait within a larger group or how one trait affects another trait in the same group.
For example, if a researcher wishes to study the disease profile of a group of senior citizens across gender, age, or socioeconomic lines, quota sampling may be ideal. The main disadvantage of this technique is that is limited to the traits that you’re studying. Other factors within the sample may be over- or under-represented, and therefore, the scope of your results will be limited. Researchers should be careful about generalizing traits outside the actual study to the larger population.
As the name implies, judgmental sampling occurs when researchers use their own judgment to select a sample based on personal knowledge and expertise. This obviously produces a bias in the sample, but judgmental sampling can be useful when studying very specific groups within the population.
For example, if a researcher is collecting insights on patients suffering from a rare disease, it would make more sense to find those individuals directly. The advantage of judgmental sampling is that you question the exact type of person you’re seeking for your research project. The major disadvantage is that you will most likely introduce human error and researcher bias into the results. As a result, it would be unwise to make generalizations to a greater population based on the results of a judgmental sample study.
Once you’ve determined the goals and objectives of your market research project, you can make a smarter decision about which type of sampling methodology to use. As you can see from this primer, you need to balance your requirements for precision against the cost and time requirements of each sampling method.
If you’re seeking quantitative results, it’s best to use one of the probability sampling methods or hybrids. If you are looking for qualitative information, one of the non-probability sampling methods may deliver the information you need at a much lower cost and time investment.
Working with a market research expert can help you better match your research requirements to the research sampling method that will give you the greatest return on your investment.
Jim Whaley is CEO of OvationMR and posts frequently on The Standard Ovation and other Industry Blogs.
Ovation MR is a global provider of first party data for those seeking solutions that require data for informed business decisions. Ovation MR is a leader in delivering insights and reliable data across a variety of industry sectors around the globe consistently for market research professionals and management consultants. Visit: https://www.ovationmr.com