Hard as it is to reach people and persuade them to start your survey instrument, you’re not done at that point. Respondents drop out as they move through the survey, which we will refer to as attrition.

If attrition afflicts them early, you collect only dribbles of useful information. If they submit a partial survey in which they went further, you have gained more, to be sure, but you would like them to have completed the survey. It is possible they will return to the survey later, but my sense is that relatively few do so (I am unaware of hosting software that reports inactive time, only elapsed time.).

I have encountered several causes of attrition and would like to suggest possible remedies or ways to lessen the dropout rate.

Too long a survey: The number one culprit, I believe, is a survey that simply wears out the respondent. “I’m tired of answering these endless questions! I’m outta here.” The prevention isn’t hard to imagine; shorten your survey so that the time you expect from respondents is reasonable.

Too much work for a question: Especially if survey fatigue has already crept up on a respondent, then a question that has 17 selections to rank or four columns in a six row table to be filled in will crack them and they will flee. When they’re at the tipping point, complexity shoves them over the cliff. Here, you might follow the ideas we have discussed for slimming down a complex question.

Too difficult a question: A question that asks the respondent to calculate something or to research a fact that they do not have readily at hand may cause them to punt. Do not ask them to fill in the ratio of partners to associates because innumerate people dislike that intimidating mathematical hurdle and they may throw in the towel. Do not ask them to tell you what percentage of their paralegals are certified paralegals, because they don’t know that figure off the top of their head and its answer would require reviewing numerous resumes or asking numerous questions. Avoid math questions; ask for the components and calculate the ratio yourself. Avoid research questions by asking for estimates or by probing for more precision with your thank-you email, ideally using a canned email.

Too invasive a question: People do turn in survey responses that ask about their sexual preferences or about discrimination and harassment, for example, but in a legal industry survey if you ask about the number of transgender personnel in a law firm, a person might rebel and quit the survey. If you ask for their age, you had better explain why you want that private item, how you will use it, and what you will do to protect it from disclosure. In a compensation survey, they know what you are about and will provide their base and bonus figures, but in a survey about causes of stress in law departments, to ask about income could be a triggering question for being too intrusive. Avoid sensitive questions unless you explain clearly why the information is so important and how you will preserve their confidences. Otherwise, a respondent may rebel right there and add another to your attrition list.