To Sample or Not to Sample?

In customer and market research this is a no-brainer. You really have no choice given the size of your target population – thousands, millions.

In employee research the same applies. However, when you are conducting an employee survey for one organization the game changes. As a general rule, I do not recommend sampling. Now, if you have a large organization (10,000+ employees) and you are only looking for an overall general answer then sampling is the way to go. But this is typically not the situation. Most organizations that conduct an employee survey want to know details. They want (need) to be able to break down the results into meaningful chunks in a statistically reliable manner.

Why would you sample in the first place? Sampling saves time and money. These are both good reasons and if you can meet your objectives only using a sample then it is an excellent approach.

Why not sample? There are two types of reasons not to sample – statistical and political. Ironically, the better job you do statistically the bigger political problem you may create. When organizations decide to sample, they usually look at the total number of employees that they have (say, 10,000) and figure that they only need 566 respondents. Now, this is true if you are only going to look at the overall results. But what they end up doing is cutting the data by several different demographics and they fail to include this in their sampling equation. Some of their data cuts are based on only a handful of respondents.

For example, if you have 10 departments of 1,000 each a comparable sample would be 375 for each department or 3,750 overall. If you want to split each of those 10 departments in male/female splits (assuming equal numbers) you would now need 273 for each gender within each department or 5,460 overall.

Another issue concerns response rate. If you need 5460 respondents (your "net" sample) and you only get a 75% response rate then you need to invite 7,280 employees (your "gross" sample). Well, now you are inviting nearly three-quarters of your organization anyway.

So, what's the political problem? "How come you didn't choose me? Most of my buddies got surveyed. Isn't my opinion important?" Now you have to explain why you chose so many but not them. If you had stayed with the original sample size of 566 employees, most wouldn't even know what was going on until you reported the results.

Since most of my clients want to be able to cut their data, sometimes all the way down to the work group level, I recommend surveying everyone. Sure, you never get a 100% response rate but you are inviting all to participate. That, in and of itself, is worth some positive political capital.

About the author: Ray Seghers is an organizational improvement consultant specializing in employee engagement and customer loyalty surveys as well as employee commitment research.
Blog URL:
Business URL:


Post a Comment

Back to TOP