People File & Sampling

When doing a survey bi-yearly, monthly or even daily you want to avoid inviting all of your employees every time: doing so will lead to survey fatigue and lower responses each time. We can help you spread out these employees over multiple surveys, doing so allows you to build a steady trend while also avoiding aformentioned problems. We can help set up the right sampling method for your project using our automated sample generator.

Cleaning the employee file

For any sample we draw we first clean up the employee file. Even when we decide to invite everyone from the employee file, while it may look like there’s no sampling going on at all, we actually draw a 100% sample. This means that we clear out any e-mails in the e-mail field that are invalid (i.e. lacking an @ symbol) and deduplicate any double email addresses. The sample generator randomly chooses which double email it will use to invite, so it’s better to have a manual look at potential duplicates before delivering the file to us.

Sampling and exclusion

The first thing we decide when drawing a sample is the percentage drawn. Our default method is to look at the total records available in the employee file, and then draw x% of that. Let’s take an example of company ACME, who delivers an employee file of 100 persons to us and wants to invite one half in March and one half in September. For March we draw 50% of the 100 persons and end up sending out 50 invites.

You might assume that for September we will do the same thing, however just taking 50% again could mean persons from March will be drawn a second time. To avoid this from happening, we include an exclusion period. In this case we would add the exclusion rule that anyone invited 6 months earlier will not be included in this second sample draw.

Added complexity

In the previous example we assume that the ACME employee file has been static. However in reality employees will have left the company, and new employees will have joined the company since then. If we take the example and say 10 persons left and 10 new persons joined ACME, then we now have 60 persons available who have not yet participated in the survey. If we were to draw 50% again, 10 employees would not receive the survey. In this case we can opt to swap the methodology around: first we exclude everyone who already received the survey and then we invite 100% of everyone who is left.

Including segmentation

To prevent a random sample from, for instance, drawing exclusively persons from one country, we can tell the sample generator to use segmentation to draw the sample on. If we for instance use country segments, the samplegenerator will group every country apart and draw seperate samples on them. In this case if the original employee file had 50% employees from country A and 20% employees from country B, then the sample will also consist of the same percentages of employees from both countries. One thing to watch out for is that the segments do not become too small. A sample of 10% on a country that only has 2 employees in it will be 0.2 employees, rounded down to 0. In this case it would be better to do the segmenting on a larger segmentation, such as region.

Was this article helpful?

Related Articles