How can we help you today?

How many responses do you need for your survey results to be statistically relevant?

Of course you’d love to have a 100% response rate for every survey you run. This would allow you to know exactly how every employee felt.

However this is an unrealistic expectation. So what is realistic?

Obviously the more responses you receive for a survey the better, because it gives you confidence the results are representative of your entire organisation. But just how many responses do you need to ensure your results are representative of the wider group, 50%, 60% 70% or higher?

Well this depends on the size of your organisation and the number of responses you need may surprise you. This is because of a statistical method called ‘Sampling’.


What is sampling?

Sampling is a way to use the results from a proportion of your employees to represent the views of your entire workforce. The purpose of using sampling is to judge if you have reached a point where your results become ‘statistically relevant’. Sampling helps you to understand how your entire workforce would score based on a smaller proportion of people responding.


What is statistical relevance?

It is when the number of responses you receive reaches a point where you can be confident they represent the views of your entire workforce. At THI we classify data as statistically relevant when they meet our sampling criteria i.e. the chosen thresholds for confidence interval and confidence level. 


What sampling criteria do we use?

When looking at sampling there are two key terms, confidence interval and confidence level.


Confidence interval

  • This is the margin of error used in the sampling calculation.
  • At The Happiness Index we have chosen to use a confidence interval of 3.
  • This means the margin of error in the sampling calculation will be plus or minus 3%.


Confidence level

  • This tells you how confident you can be in your results.
  • At The Happiness Index we have chosen to use a 95% confidence level.
  • This means 95 out of 100 times a survey is asked, the results will fit within the chosen confidence interval.

Here's an example

  • Your survey average is 7.6.
  • The chosen confidence interval is 3.
  • The chosen confidence level is 95%.
  • This means 95 out of 100 times you run the survey you can be confident your survey average would be between 7.4 and 7.8 (e.g. 7.6 plus or minus 3%).


This table shows how many responses are required for your survey results to be statistically relevant based on a confidence interval of 3 and a confidence level of 95%.


Why do you need a lower percentage of responses as the number of employees increases?

You may have noticed in the table above the larger the number of employees you have the smaller the percentage of responses required becomes.

This is because as the number of responses increases the level of variation between employees starts to decrease. This means the accuracy increases and you don’t need as many responses.

Here’s an example

Imagine you’re an organisation of 1,000 people and you asked 5 employees to rate how happy they are at work on a 1-10 scale. How confident would you be that their responses would represent the view of your entire workforce? Probably not that confident, as you could have selected 5 people who have views that are quite different to the overall workforce, e.g. 5 people who are particularly unhappy at work.

Now imagine you asked 500 people, you’re going to be more confident that their views are representative of the entire workforce. This is because there will be less variability in their scores and therefore you can be more certain their results are accurate.