Survey fatigue: how often can you ask the same people?
Every survey you send makes the next one a little harder to fill. Here is what the research says about asking the same people again and again, and how to set contact rules that protect your response rate.
By SurveyLane · The team building SurveyLane
Your survey does not land in an empty inbox. It lands next to the onboarding survey product sent last week, the support ticket rating from yesterday and the engagement pulse HR has scheduled for Monday. The respondent does not see three teams with three good reasons. They see one company asking again. Each request spends some of their goodwill, and the next survey pays for it, whether it is yours or a colleague's.
Two kinds of survey fatigue
People use the term for two different problems, and the fixes differ.
The first is fatigue inside one questionnaire: by question 25 people tick the same column down a grid or close the tab. That is covered in the post on survey length and dropout.
The second is fatigue across surveys: the person has been asked so often that the next invitation goes unopened. Rewriting question 25 does nothing when they never reach question 1. This post is about that kind.
The evidence here is thinner than for questionnaire length. In a 2022 review in the Journal of Interdisciplinary Studies in Education, Barry Fass-Holmes concluded that survey fatigue "has not been rigorously defined". Be suspicious of precise rules of thumb, including the ones below.
What the Wesleyan experiments measured
The most useful study is old and small, and still unusually clean. Stephen Porter, Michael Whitcomb and William Weitzer ran two randomised experiments at Wesleyan University and published them in 2004, in issue 121 of New Directions for Institutional Research. They deliberately sent some people more surveys than others and compared what happened next.
In spring 2001, the 649 seniors were split at random into two groups. One group got a seven-page questionnaire about their college experience in late March and April. Then every senior got the eight-page Senior Survey from late April into May. Seniors who had not been asked before responded at 67 percent. Seniors who had just been sent the first survey responded at 57 percent. One extra request, sent almost directly before, cost 10 percentage points.
The second experiment ran in the 2002-2003 academic year. First-year students were split into four random groups of 144 and got between one and four web surveys over the year. For the dining services survey in late November, the response rate was 68 percent among students with no earlier survey, 58 percent after one and 46 percent after two. A drop of 22 points, with all three surveys inside a single semester.
Back-to-back requests hurt most
There was a fourth survey in March. Students who had received three surveys in the autumn responded at 47 percent, against 60 percent for students who got only this one. That is still a 13-point loss. It is also clearly smaller than the 22 points in November, when the requests sat closer together.
The authors concluded that fatigue is strongest when surveys run back to back and seems to wear off over time. A survey from the previous semester may barely touch the next one.
So clustering matters more than the yearly count. Four surveys over twelve months feel nothing like four in six weeks, though both read "four a year" in a planning sheet.
Relevance matters more than the count
One group's response rate went up on its second survey. That survey was about campus dining. It was short, and it was about something every first-year student dealt with daily. The orientation survey, about something the students had just been through, also left little measurable fatigue behind. The damage came from a national alcohol and drug questionnaire that the authors describe as long and personally intrusive.
An earlier study at the U.S. Air Force Academy, which Porter and colleagues cite, points the same way. Of the cadets asked, 97 percent felt at least somewhat oversurveyed and 48 percent said "yes, definitely". Asked what that meant, they described frequent surveys that felt irrelevant to their daily life.
This matches the leverage-salience theory that Robert Groves, Eleanor Singer and Amy Corning published in Public Opinion Quarterly in 2000. People weigh several features of a request, such as topic, sponsor and burden, and those features count differently for different people. A survey about something the respondent cares about can outweigh the fact that it is the third one this month.
So "send fewer surveys" is the wrong rule. A better one: don't send surveys people have no reason to care about, and never send them back to back.
The people who keep answering are not everyone
The Wesleyan response rates never collapsed. No survey in the second experiment went below 44 percent, however many requests came before it. The authors suspect a core of reliable responders who answer whatever you send.
That should worry you. If heavy surveying strips away everyone except your most cooperative people, the response rate levels off and looks stable while the people behind it drift further from the population you want to describe. Loyal customers keep answering, and the ones quietly leaving stop. Engaged employees fill in the fifth pulse. The disengaged ones tuned out at the second.
In several of the study's tables the drop was bigger for some subgroups, though with groups this small most gaps were not statistically significant. The post on sample quality explains why a low response rate is not the real problem and a skewed one is.
Why the pressure keeps rising
The 2004 authors already worried that cheap web tools would let anyone send a survey. Today every support desk and HR platform has one built in. Pew Research Center reported in February 2019 that the response rate on its telephone surveys fell to 7 percent in 2017 and 6 percent in 2018. Nobody can pin that on one cause, but when everyone asks more, each request is worth less.
Expectation counts as well. Porter and colleagues cite a 1998 experiment by Apodaca, Lea and Edwards in which just telling people they would be contacted again several times over the coming years lowered the response rate by 5 percentage points. People react to the burden they expect, too.
Make a calendar of every survey that reaches your audience
You can't manage a frequency you can't see. Start with one list of every survey that reaches the same people, whichever team sends it.
For customers that means ticket ratings, NPS, onboarding and churn surveys, research recruitment and event feedback. For employees: the engagement survey, pulses, onboarding and exit surveys, training evaluations and whatever managers send on their own.
Note who receives each one, what triggers it and who owns it. You will find overlaps nobody planned, like this one: a customer opens a support ticket, gets a ticket rating, gets the quarterly NPS survey that same week and then an invitation to a research interview.
Set a contact rule per person
With the calendar in front of you, add a rule that works per person instead of per survey: a minimum gap between any two survey invitations to the same individual. Thirty days is a common starting point for customer surveys. Treat it as an assumption to test against your own data. It is not a research finding.
When two surveys collide, one waits. A workable priority order: anything the person asked for, then transactional surveys about something that just happened, then periodic surveys, then everything else. A research team that wants "just a few customers" should not get to skip the rule because their survey is short.
Respondents also don't separate your survey from your reminder. An invitation plus two reminders is three contacts, so count them in the rule. The post on getting invitations delivered explains why every extra email also costs sender reputation.
Sample instead of asking everyone
The simplest way to ask each person less often is to stop asking everyone every time. Say you have 20,000 customers and need 400 good responses for the precision you want. Sending the survey to all 20,000 every quarter wastes goodwill you will need next quarter.
Draw a fresh random sample each round and rest the people you asked last time. Over a year each customer is asked once instead of four times, and you still get four measurements. The post on how many respondents you need shows how to work out the number.
Small populations don't have that option. A team of 40 cannot be sampled, so every survey reaches all of them and coordination matters even more.
Trigger-based surveys need a brake
A rating after every support ticket sounds reasonable until a customer with a messy problem opens six tickets in two weeks and gets six identical surveys. The person you most need to hear from is the one you annoy most.
Give every trigger a suppression rule: survey only the first ticket in a given period, or only the ticket that closes the case, and never if the person had any survey invitation recently. Conditional logic routes the questions inside a survey. Whether to send the survey at all is a job for the contact rule.
Show people what the last survey changed
The cadets complained about irrelevant surveys. Relevance has a second side: whether anything visibly happens after you answer.
If people fill in your engagement survey and hear nothing for a year, the next invitation reads as a reminder that the last one went nowhere. A short note before the next round changes the deal: "You told us X, so we changed Y. We didn't do Z, and here is why." The post on closing the feedback loop shows how to do that per respondent instead of in a company-wide memo.
Watch your own numbers for the first signs
Your own data shows fatigue if you look by contact history instead of by survey. Split each survey's response rate by how many invitations the person got in the previous three or six months. If the rate drops with every extra invitation, you are seeing the Wesleyan pattern in your own population. Check quality as well. More straight-lining and shorter open answers among frequently asked respondents are fatigue signs that never show up in the response rate. The post on monitoring response quality lists what to look at.
Then compare respondents and non-respondents on what you already know, such as tenure, plan and region. If that gap widens each round, your trend lines now describe a different group. Fix the contact rules before you touch the questionnaire. These people stopped before they ever saw it.
Frequently asked questions
How many surveys a year is too many?
There is no research-backed number that holds for every audience. The Wesleyan experiments show that two or three surveys in one semester already cut response rates sharply, while surveys months apart did much less harm. Start with a minimum gap per person, such as 30 days for customers, and adjust it based on how response rates in your own data change with contact frequency.
Do reminders count as extra contacts?
For the respondent, yes. Someone who receives an invitation and two reminders has been asked three times. Count reminders in your contact rule, keep them to two at most, and stop them as soon as the person responds or opts out.
Is it better to survey everyone less often or a sample more often?
For most purposes, a random sample each round. You get regular measurements, each person is asked less often, and you can rest people who were asked recently. In small populations such as a team of 40, sampling does not work, so coordinate the calendar and cut surveys that do not lead to decisions.
Does a short survey cause less fatigue?
Short helps, but relevance helps more. In the Wesleyan study, a short survey on a topic students cared about barely affected later response rates, while a long and intrusive questionnaire caused most of the drop. A short survey on something the respondent does not care about still uses up goodwill.
