A pulse survey is a short, recurring employee survey, usually 5 to 15 questions on a monthly or quarterly cadence, with the wording held identical between waves so the scores form a real trend. It complements the annual engagement survey; it does not replace it. This page covers the cadence, the questions, and the part most guides skip: how to tell a genuine trend from sampling noise.
By the Wavefield Research team · Published Nov 4, 2026
Employee listening done properly uses two instruments with different jobs. The engagement census (20 to 40 items, once or twice a year) measures levels: where the organization stands on the full model of engagement, team by team. The pulse check survey (5 to 15 items, monthly or quarterly) measures change: which of a handful of indicators moved since last wave. The census is a map; the pulse is a smoke detector. Our full guide to the census side, question set and response-bias math included, is the employee engagement survey page.
The reason organizations bother with either is well documented. Gallup's Q12 meta-analysis (11th edition, covering 183,806 business and work units with 3.4 million employees) reports a 23 percent median difference in profitability and a 78 percent difference in absenteeism between top-quartile and bottom-quartile teams on engagement. None of that proves your monthly pulse causes anything, and a vendor page that says otherwise is selling. What the evidence supports is more modest and still worth the effort: engagement relates to outcomes strongly enough that noticing a decline early, while the cause is one quarter old instead of four, has real value.
One design rule does most of the work: the employee pulse survey is a tracking instrument, so the core wording is frozen at wave one. Change a question and the trend line breaks at that point, permanently. Rotating topical questions belong in designated rotating slots, never in the core.
Monthly or quarterly suits most organizations, and the deciding constraint is not how often you can send a survey. It is how often you can visibly act on one. Each wave spends trust, and the only thing that refills trust is people seeing something change because they answered. A pulse program on a cadence faster than its action loop trains the workforce to ignore it, and the decay shows up as quietly collapsing response rates. The mechanics of that collapse, and what response rates are normal to begin with, are covered in survey response rates.
The default. Four waves a year is enough to see direction, slow enough that every wave can be followed by visible action, and light enough that nobody resents it.
For organizations in flux: fast growth, reorgs, a turnaround. Worth it only if someone owns the monthly action loop. Twelve waves of unactioned data is a fatigue machine.
Almost always a mistake outside short exceptional periods. Week-to-week movement on a team-sized sample is mostly noise, and answering the same question 52 times a year is a chore, no matter how friendly the widget.
Length: 5 to 15 questions, under five minutes. A workable composition is three or four frozen trend items, two or three rotating topical items, and one open end. If the list keeps growing, the extra questions belong in the annual census, not the pulse.
Good pulse survey questions share three properties: concrete enough to answer in seconds, sensitive enough to move between waves, and attached to something someone will actually do. Eight questions cover the standing needs; no single wave should run more than five or six of them plus the open end. The wider employee question bank, lifecycle stages from onboarding to exit, lives on employee survey questions.
“How would you rate your overall experience of working here this month? (Very poor to Very good)”
The headline number. Its only job is the trend line, so the wording gets written once and never edited. A pulse with a rotating headline question is a stack of unrelated one-off polls.
“I would recommend this company as a place to work. (Strongly disagree to Strongly agree)”
The eNPS-style loyalty item in agreement form. Track the favorable share (agree plus strongly agree) rather than a mean score; movements in the favorable share are easier to explain to managers and harder to fake with scale drift.
“I have what I need to do my job well. (Strongly disagree to Strongly agree)”
A resources item in the tradition Gallup's Q12 established: concrete, about daily experience, answerable by anyone. Items about abstract pride or alignment move slowly; items about tools and clarity move while you can still act.
“My workload over the past few weeks has been... (Far too light / About right / Far too heavy, 5 points)”
The earliest burnout indicator a survey can carry. Read it by team, never only in aggregate: a company average hides the one group running at 140 percent.
“My manager checked in on my progress and blockers recently. (Yes / No)”
A behavior item, which is the trick: people report what happened more reliably than how they feel about it. A falling yes-share precedes a falling engagement score by about a quarter in most programs.
“I know what I should be prioritizing right now. (Strongly disagree to Strongly agree)”
Clarity is the cheapest thing leadership can fix and one of the fastest items to respond when they do, which makes it a fair test of whether the pulse program itself works.
“What is one thing we should change before the next check-in?”
One open end per pulse, phrased as a single priority. It converts the survey from a scorecard into a suggestion channel, and the themes in these answers explain the score movements the closed items can only detect.
“Is there anything this survey should have asked about?”
The escape valve. Run it quarterly rather than every wave; its real value is telling you when the fixed instrument has drifted away from what people actually care about.
Every pulse survey guide shows a dashboard with a line going up. Almost none of them say when a movement in that line means anything, and on the group sizes involved in employee research, that is the whole question. Sampling error on small groups is brutal: AAPOR's margin of sampling error explainer puts the margin for a sample of 200 at about plus or minus 6.9 percentage points, and notes the margin falls sharply between samples of roughly 100 and 1,000. A 60-person company whose favorable score moves from 72 to 68 percent has learned approximately nothing; the movement is comfortably inside the noise. (Strictly, those figures describe probability samples; a pulse is usually a census with nonresponse, where the same small-numbers caution applies through a different mechanism: a handful of people changing their answer, or skipping the wave, moves the score by points.)
Three habits keep the readings honest. First, judge trends on three or more waves, never on one step: a single-wave dip is a coin flip, three consecutive declines are a signal. Second, watch the response rate next to every score: a favorable score that rose while response fell may only mean the unhappy people stopped answering. Third, respect group size when drilling down. Team cuts below about 20 responses move for arithmetic reasons (one person is 5 points), and below the anonymity floor of roughly 5 they should not be shown at all. The same small-numbers logic governs sample sizes generally; the sample size calculator makes the margins concrete for any group size you care about.
And one habit protects the trend itself: never rewrite a core question to “improve” it. A wording change moves scores on its own, so the improved question starts a new trend line and orphans the old one. If a question must change, run old and new side by side for two waves so the series can be spliced.
What kills a pulse survey program is silence after the results: people answer honestly twice, watch nothing happen, and conclude the survey is theater. The questions are rarely the problem. The loop that prevents this fits in four steps, and the calendar should show all four before wave one ever sends.
Within two weeks of fielding: here is what you said, here is the one thing we are doing about it, here is what we are deliberately not doing and why. The second half of that message buys more trust than the first.
Team-level results go to team leads, above the anonymity floor, with last wave beside this wave. A manager who first sees their team's decline in a company all-hands has been set up, and will treat the program accordingly.
One fixed meeting cadence, one tooling fix, one clarified priority. Small actions that trace plainly to the survey beat grand initiatives that do not, because the point of acting is partly to prove the channel works.
Any question whose answers have never changed a decision is costing attention for nothing. Cut it. A shorter pulse people believe in beats a thorough one they have learned to click through.
On tooling: continuous listening suites price per employee per month, which is the right model for always-on programs and an expensive one for a quarterly pulse. We run pulses as tracker waves instead ($99 a wave, instrument held constant); the honest comparison of both models is on employee survey software.
A pulse survey is a short, recurring survey, usually 5 to 15 questions, sent to employees on a fixed cadence such as monthly or quarterly. The questions stay identical between waves so the scores form a trend line. It complements a fuller annual engagement survey rather than replacing it: the census measures levels in depth, the pulse tracks change between censuses.
Early warning and trend tracking. An annual survey tells you how people felt months ago; a pulse catches a falling workload score or a cooling manager-support score while the cause is still fresh and fixable. The purpose is never the score itself: a pulse program that produces numbers but no visible actions trains people to stop answering.
Monthly or quarterly for most organizations. Match the cadence to your acting capacity: each wave should land after the previous wave's findings produced something visible. Weekly pulses suit almost nobody; response rates decay, and week-to-week movement in a small company is mostly sampling noise anyway.
Five to fifteen, answerable in under five minutes. A useful composition: three or four fixed trend items, two or three rotating topical items, and one open end. The fixed items never change wording; the rotating slots are where this quarter's question about the reorg or the new tooling goes.
Length, cadence, and job. An engagement survey (census) runs 20 to 40 items once or twice a year, aims for everyone, and measures levels across the full model of engagement. A pulse survey runs 5 to 15 items monthly or quarterly and measures change on a handful of indicators. Programs fail when one tool is asked to do the other's job: a 40-item monthly survey, or an annual-only program that notices problems a year late.
They should be, and anonymity needs a reporting floor to be real: no group result shown below a minimum of roughly 5 responses, because in a 4-person team an anonymous answer is not anonymous. Say the floor out loud in the invitation. Confidential-but-identified designs (a vendor holds identities, managers see groups) also work, but only if that is stated plainly, never discovered later.
Related: employee engagement surveys · employee survey questions · employee survey software · survey response rates
Build the instrument once, re-field it every wave for $99 with the wording held constant, and read the trend on honest bases. Free to build and test.
Build a pulse survey free