Whether it's online or offline, if we're not doing focus groups, we're usually using surveys to collect Primary Data. Surveys give us the opportunity to draw conclusions about different groups of consumers by studying a small statistical sample of the total consumer population. The "sample" is the key. A sample is usually defined as a segment of the population selected for our research that will represent the larger population as a whole. Whether or not a sample is good enough to make observations with, depends on how we've designed it.
Designing a sample is a three step process:
1) Decide who you are going to survey
2) Decide how many you are going to survey
3) Decide how you are selecting the participants in the sample
Let's examine each of these three steps a little more.
1) Decide who you are going to survey
First, you have to decide who you are actually going to survey. In more statistical terms, we are asking "what is the sampling unit"? Any group of people can be used as a sampling unit. What I mean here is children, adult women, men, etc. Your sampling unit should be determined by the target groups in your survey, and the data you have about your target customers. If you don't know who your target customer is, then you have some research to do first. Choosing the wrong sampling unit will waste your time and your money. It will give you data that you cannot use, because the results from that group will be irrelevant.
2) How many should be surveyed (what is the sample size)
When we are asking ourselves "how many should be surveyed", what we are saying here is "what is the sampling size?" Sample sizes that contain more people usually give us increased accuracy. For you statistical junkies, there are certain facts of mathematical statistics that describe this, such as the law of large numbers and the central limit theorem. To keep this simple for our discussion, larger samples will give you more statistically reliable results than smaller sample sizes. However, larger sample sizes will cost you more money. Do not assume that you need to attempt to sample an entire population segment (which would take forever, and be almost impossible). Usually less than 1 percent of a population segment will provide statistically reliable results. There is a down-side to Probability Samples: cost. Depending on your Contact Method, larger samples will result in drastically higher costs. When cost is a factor, then researchers turn to Non-Probability samples.
3) How should the people in the sample be selected?
What we are asking here is, "What is the sampling procedure we are going to follow?" There are two different types of samples we can choose from: Probability Samples, and Non-Probability Samples. Probability Samples give each population member (a.k.a. a potential participant) a possible chance of being included in the sample. Because you are not sampling the entire population, probability samples will always contain margin for error. The larger the margin of error, the less trust you can place in the data you have that is supposed to represent your "population". Larger samples give you less margin of error, and less margin of error lets you trust your data more.
There are three different types of Probability Samples:
Simple Random Sample
Every member of the population has a known and equal chance of selection.
Stratified Random Sample
The population is divided into mutually exclusive groups (such as age and race) and random samples are drawn from each group. Basically, you are splitting your population into defined groups, and then sampling each of those groups.
Cluster (area) Samples
The population is divided into mutually exclusive groups (such as blocks, and they are relatively homogeneous) and the researcher draws a simple random sample of each group.
Non-probability sampling is much less expensive than doing Probability Sampling, but the results are of limited value, because the data is less reliable. Non-probability samples should be used with caution. Non-probability sampling techniques cannot be used to deduce generalizations from the sample to the general population. Any generalizations created from a non-probability sample MUST be filtered through the researcher's knowledge (and yours) of the customer population being studied.
There are three different kind of Non-probability samples:
Convenience Sample
In a Convenience Sample, the researcher selects the "easiest", most convenient to locate member from the immediate population to obtain research data from. I would consider this one of the most hap-hazard methods. There is practically nothing you can generalize about the data you obtain, other than considering it a "snapshot" of a a particular group, at a particular time, at a particular place.
Judgment Sample
In a Judgment Sample, the researcher will use his/her judgement to select the the people sampled. Immediately you have to ask...how good is their judgment in selecting a good candidate? Again, the data that you obtain from this sort of sampling is just not reliable for generalized conclusions, but it may be interesting to use as a "snapshot".
Quota Sample
Probably the worst method I can think of, a Quota Sample is simply a researcher grabing enough people to meet a quota requirement for sampling participants. Stay away from this sort of methodology. Your data is practically useless.
Never select a sampling method without taking the time (as you always should) to weigh the needs you have when collecting Primary Data. Always take into consideration your time-frame and your budget, and always try to be as objective as possible when you are evaluating the Primary Data you have obtained.
Follow @macdailybites
0 comments:
Post a Comment