There are many types of sampling methods available in a market researcher’s toolbox.  Each method has its pros and cons.  Some methods, much more so than others, are suitable for specific types of marketing research.  Some sampling methods are truly random in that each member of the specific population has an equal chance of being selected; this chance of selection can be represented as a probability (i.e. 1 in 1,000 chance of selection) and is therefore known as a probability sample.

1)  Simple Random Sampling.  This is one of the simplest and most commonly used forms of probability sampling.  There are multiple ways to approach the random selection process to ensure that each member of the population has an equal chance of selection.  The key is that the respondents are chosen randomly.  Although there is no guarantee, when you choose a random sample of participants from a pool, it should be representative to the makeup of the population as a whole.

Example: A school district is hoping to learn more about their students.  Their student body consists of 10,000 people and they want to send survey invitations to 500 of them.  The total sample of students is randomly sorted and the first 500 students are selected to be surveyed.

2)  Systematic (Random) Sampling .  The alternative to choosing the respondents randomly is choosing them systematically.  Systematic sampling involves developing a specific sequence in which respondents in a sample are chosen.  Using a random starting point, you can use an interval to sweep through a sample list.  This process ensures a thorough and even sampling.

Example: A medical center had 2,000 unique patient visits over the last three months.  They would like to survey 400 of them to learn more about their satisfaction and impressions of the facility.  They create a systematic sampling method by randomly selecting an interval, say five.  They then start at a randomly selected point in the sample and choose every fifth record of the sample until they have 400 patients.

3)  Stratified Sampling.  This sampling method is based on strata, also known as subpopulations.  The population is broken into strata based on specific characteristics.  After the sample is broken up, strata can be randomly selected for conducting market research.

Example: A senior living community developer is looking to institute new activities and opportunities for the members of their communities.  They are quite confident that those in the older age ranges will react differently than those in the younger age ranges.  They take their population of 2,000 residents and split it into a group of age 62 to 69, a group of those 70 to 79 and a group of those 80+.  By randomly surveying each stratum, it ensures that there are enough respondents from each age group to make meaningful comparisons.

4)  Cluster Sampling.  This  method is similar to simple random sampling; it is seen as an alternative for use in certain circumstances (in which case normal methods of surveying people in all different areas is inconvenient).  The population is divided into mutually exclusive clusters.  This method relies on assumption that each group is quite similar.  Certain clusters are then randomly selected and the people within the cluster are surveyed.

Example:  A school district is hoping to conduct qualitative research with the community to learn more about impressions of programs they are looking to institute.  They separate the district into 30 major neighborhoods.  The school district randomly selects five of the 30 neighborhoods for the research.  By doing this, the sample still remains randomized, while cutting administrative, logistical and travel costs.

Probability sample methods in most circumstances are the way to go with market research, unless you are hoping to utilize the benefits of a specific non-probability sampling method, or you are limited by things such as budget  or access to a full population list.  These methods allow the market research to confidently state the results are representative of the population (to a certain margin of error, of course). 

Look for a blog post later this week about the non-probability sampling methods!