How to define the sample of your marketing research
Knowing how to define a survey sample doesn’t have to be a complicated process, but it’s essential to pay attention to a few steps summarized in this article.
What is a research sample?
A survey sample is a term that represents the portion of the population interviewed in a survey.
The sample is a cut that correctly respects the population’s proportion to represent it. It is essential to know the concept of a research sample because it is practically impossible to interview the entire population.
Imagine that you want to survey the target audience of females over 50 years old throughout the United States. As you do not interview every American man, you limit the survey sample to a certain number of people who meet that criteria.
This example will define how many men to interview through the precision process. In addition, you will understand and determine which samples this group must meet so that your result will be statistically significant.
How to define your sample amount
The first step is defining your criteria. Following the example above, the total population of women in the United Stated above 60 years old is 41.65 million.
Let’s pretend Google wants to understand how often females 60+ years old in the US access the internet. It is impossible to count 41.65 million answers to our research. Hence, we need to come up with a sample number that will be statistically significant to represent the behavior of the total population.
To define the sample size, there are at least three factors we need to consider, based on your research:
Factor #1: The confidence level
Factor #2: Estimated variance
Factor #3: Allowable error margin
The sampling formula
n = Z²(pxq)/e²
You might be thinking: “ay, math? No, thanks…” or “what is this crazy formula?”, which honestly, I relate with you. I am a marketer, and I am not a fan of crazy formulas; however, the importance of this one is high for marketers not to master it.
The N stands for the sample, the actual number that I want to find in the end as my total research sample size.
Z represents the confidence level you want in your research. This confidence level usually varies between 90%, 95%, and 99%, and each percentage has a corresponding Z score which is the actual decimal to plug into your formula (1.65, 1.96, and 2.58).
(p x q) is your estimated variance, where p is the percentage of any prior study you have for the same research objective, and q is a simple mathematical formula: q = 100 – p.
Last but not least, e represents the error margin for the survey. This error varies a lot based on the accuracy of your results.
Practicing
Let’s go back to our example: Google wants to understand how often females 60+ years old in the US access the internet.
Total population: 41.65 million
Z = confidence level in 95%
p = no previous survey had been done
q = 100 – 50 = 50 (when there is no survey done before, p = 50 automatically)
e = error margin in 3%
n = (1.96)²((50x(100 – 50)))/(3)²
n = 3.8416 x 2.500/9
n = 1,067.11
Out of curiosity, all percentages in this formula format are added as real numbers. The error margin is 3%, so we add the three instead of 0.03. We never transform percentages for this formula!
Now… is it correct to say Google needs to interview 1,067 women older than 60 years old in the US?
And the answer is no. There is a 0.11 in this calculation that cannot be ignored, so the final sample for this research would be 1,068 people following the criteria.
Well, out of 45.67 million women in the US older than 60 years old, considering 1,067 of them is way more realistic, right? That’s the power of sampling!
Now that you know how to create your research objective and how to define your sample, what do you think is the next step to proceed with your marketing research?
Share your insights below!