How To Write An Employee Survey
As with many other professions, survey writing is a craft that can take years to develop, even though it may look easy. Before you design your own employee survey, you really need to examine why you have decided not use a professional research company. Does your company also write its own advertising or do you use professional help? What about your HRIS system? Are you using an external system or did your company program an internal resource? Are you sure that you can design a survey that will lead to effective management decisions? Keep in mind that an employee survey is one of the ways in which you present your organization to your employees and, separate from questions of data and analysis, a professionally-designed survey is more likely to encourage participation and build trust than will a homegrown approach.
A second key question concerns anonymity. Keeping your employees' thoughts and opinions anonymous is absolutely essential if you want them to give you honest feedback. However, if you are collecting the data yourself, what guarantee can you give your employees that their answers will be completely anonymous? Remember that there are two equally important components to anonymity - (1) are your employee results actually anonymous or is it possible to link their findings to some personally-identifiable information and (2) do your employees believe the results are anonymous? The perception of anonymity is the only factor that counts when your employees are filling out their survey. If they decide not to be honest or to report only what they think you want them to say, your survey results will be dangerously flawed but you won't know by how much!
Thirdly, survey results are only as good as the analysis conducted to examine themes, look for the data patterns and draw meaningful conclusions and recommendations. How much experience do you have with data analysis? If this is the first time you are analyzing data of this type, are you comfortable drawing the necessary conclusions? Remember, if you use your own survey, you will not have any normative data to help put your findings into context. How are you going to know if your results are good or bad?
If you have answered these questions and are still determined to write your own survey, here are some guidelines that should make the process a little less risky:
These guidelines will help you create your own employee survey if you make that choice. In the end, though, the question remains - why don't you leave this task to the professionals?
- First of all, gather detailed information on the kinds of information you need to have. What are some of the recent changes that have been happening within your organization? What has prompted the need to conduct a survey now? What findings are absolutely critical to know? Try projecting out to designing an Action Plan - what would be helpful for you at that stage of the process?
- Recognize that a well-designed survey is always much more than a collection of questions. Take that issues you have identified and group them according to themes or topic areas. Create an outline of the survey before you try to write any individual questions. Question flow is extremely important, with one set of questions naturally leading to the questions that follow. Poorly-designed surveys often jump around topics randomly and this can be very frustrating to the people who are answering your survey.
- Think about how long you expect your survey to be. Very short surveys may be quick to do but lack a lot of depth - if your employee survey is too short to help you decide what to do, what benefit did you get from it? Organizations are very complex, as are the people who work in them, and it is not realistic to get a comprehensive picture of your organization in a very short survey. At the same time, though, if a survey takes longer than about 30 or 40 minutes, you run a high risk that many employees will abandon the survey midway through.
- An important principle in survey design is to have at least one measure (what statisticians call the "dependent variable") that is meant to be an overall summary of the findings. This question should be early in the survey, so the responses to this question cannot be influenced by any follow-up question. In employee surveys, overall job satisfaction is good example of this type of question.
- Survey design is all about scales, so you need to decide how many scale points you are going to include. There are plenty of arguments out there in favor of 4 point scales or 5 point scales or 7 point scales or 10 point scales, etc. One guideline is that the number of scale points should be in line with the level of precision you are measuring. If you include too many scale points, you run the risk of "numeracy," which means that your results will look far more precise than they actually are. Employee surveys are based almost entirely on recall and impressions, which do not lend themselves to highly precise scales.
- On the same point, most surveys will contain a mix of different scales (although they will likely have the same number of scale points) because it is almost impossible to force fit a range of topics into one solitary scale, such as an "agreement" scale. A range of different scales not only gives you more accurate results, it also makes the survey more interesting for those taking it!
- Another consideration is whether to include any freeform text questions and, if yes, how many? Giving your employees the opportunity to express their thoughts and feelings in their own words can be extremely valuable. However, inviting them to do so too often can lead to frustration and cause them to abandon the survey. One of the goals of all surveys is to find the right balance between quantitative rating questions and open ended freeform text questions.
- One open ended question you should always avoid is a final "catch-all" question, like "is there anything else you would like to add?" It does not say much about your survey if you need your employees to tell you what topics you have missed!
- Finally, if you have designed a survey "from scratch," you really need to test that survey before you launch it. There are two ways to test surveys - cognitively and in-field - and you should be doing both. Cognitive testing means that you ask a group of people to answer your survey as naturally as possible, then you go through each question with them one-by-one. What you are looking for is: (1) did they understand the question, (2) did the choices provided fit with their experience and (3) did the answers they give accurately reflect their experience for that question? This careful testing can be very worthwhile in identifying problems and question that do not ask what is intended and giving guidance on how to fix those problems. In-field testing means that you make sure the survey works as it is intended to work - for example, if you are using an online survey, could your respondents log on easily? Were the survey instructions clear? If you are using paper surveys, did your respondents know what to do once they have finished their surveys? The purpose of the in-field testing is to see how if all of the "nuts and bolts" are working well.