For the sake of argument, let’s take a case where I ask 2 customers whether they were comfortable with the air conditioning. One says Yes and one No, could you confidently deduce that there is a 50% satisfaction level? The obvious answer would be no.
Taking the same argument ahead I survey 100 customers and 50 say they are happy with the air conditioning I come up with the same 50% level of satisfaction. But I can be more confident about what I deduce in the second case.
Putting it numerically, there are two factors that decide how many customers I should survey in order to judge satisfaction levels.
- Confidence Level: This is the amount of confidence I need in judging the satisfaction level of the customer community based on the sample I survey. 95% is usually a standard confidence level used in statistical evaluation. Which means that with the sample size I determine, I would make the correct judgment about the satisfaction levels 95 times out of 100 times that I conduct the survey.
- Margin of Error tolerable: I find as a result of the sample survey that the satisfaction levels are predicted at 80%. The 80% estimate is subject to a certain amount of error. This error can be limited to a predetermined level with a suitable sample size. In this case the margin of error band would be 4% (5% of 80%). This would mean that my estimate of satisfaction is a percentage in the band of 78% - 82%.
Understanding the level of customer satisfaction within a 5% margin of error is usually workable for most of us. If you need a better accuracy, say limiting the margin for error to 1%, your sample size will go up significantly without really giving you a substantial added benefit.
- As the tolerable margin of error decreases, the sample size goes up. This is pictorially depicted in the animation here.