what is an anonymous pulse survey? a manager's guide

you manage a team, you suspect you hear the safe version of how it's going, and someone has suggested a pulse survey. this is the guide we would have wanted: what one is, what "anonymous" has to mean for it to work, how often to ask, what to do with the answer, and how to start. we make one, so we have a view. we say where it stops.

what a pulse survey is

a pulse survey is a short survey that you repeat. a few questions, taking a minute or so, asked on a rhythm instead of once a year. the point is not any single answer. it is the line the answers draw over time: is something getting heavier, did the change you made help, is the reorg landing the way people said it would.

it does one job that a one-to-one cannot. in a one-to-one, the person across from you decides how much of the truth is safe to say. a good pulse survey lets the team tell you without anyone deciding that. it does not replace talking. it tells you where to start the conversation.

what "anonymous" has to mean

most tools say anonymous. there are two ways to mean it, and your team will be able to tell which one you bought.

anonymous by policy means the tool stores who said what and promises not to show you. that can be a sound design, with thresholds that hide small groups. but the promise lives in the settings, and settings change, admins differ, and people on your team know that.

anonymous by structure means the answer is stored without the person. in jollygig an answer is saved as the team, the check-in it belongs to, and three multiple-choice choices. no name, no email, no user id, no time of its own, and no free-text box, because a writing style is a name. there is a separate "did you answer" record, so people can see their own history, but it is never joined to an answer. you can see who is on the team and who has joined. you cannot see who answered.

the floor

structure is not enough on its own. imagine a team of three, two answering "fine" and one "not great". you know the team. the result is a thin wrapper around three people. so a result needs a floor: a minimum number of answers before anything shows. ours is five, it is fixed, and no admin setting lowers it. every breakdown, by tenure for example, needs five of its own. below that, the product shows nothing, and says so.

the idea has a name, k-anonymity (sweeney 2002): a result should describe at least k people, so nobody stands out. five is a convention, and the default in tools such as microsoft viva glint. it is not a guarantee. if all five people give the same answer, the answer is still visible, so in that case we show you the count and not the pattern. and because a manager who knows their own answer could subtract it, you do not answer your own team's check-in and are not counted toward the five.

what is left

a vendor cannot prove a negative, so here is ours. like every database record, an answer has a creation time in the database's own metadata. we never read it, export it or show it. but the people who run jollygig can technically reach the database, as with any hosted service, and someone with that access could try to pair a creation time with a participation record. access is limited to the founder and logged by google cloud. our claim is about what is stored and what the product can show you, not that it is impossible for anyone to find out. the privacy page says the same in more detail, and can my manager see my answer? is the version for your team.

how often to ask, and why we default to every two weeks

there is no tidy answer here. we looked for a study of how often a pulse should run and did not find one that settles it, and the figures that vendors quote are rarely sourced. what research there is points one way: short beats long, and a reminder helps (dillman, smyth and christian 2014). past that, it is a judgment.

ours is every two weeks, and it is a default, not a finding. you can choose weekly or monthly when you set up a team, and change it later. we lean on two reasons. it asks less of people, and a survey people keep answering is worth more than a sharper one they start to skip. and most of what a team feels, like workload, priorities and how work is handed over, does not move in a week. the second reason is our reading, not a measurement.

how it runs: a check-in opens for a few days, three questions, under a minute, with one reminder before it closes. the first one waits until five people have joined, or 72 hours after your first invitation, whichever comes first, so your first read is not an empty one. if fewer than five answer, nothing is worked out at all. nothing yet means nothing yet, not nothing wrong.

what to do with the result, and what not to

a result describes the team. it is a pattern, something like "a good share of check-ins reported feeling stretched about workload", with a note on how sure we are. it does not describe a person and it should not be used to find one.

that is a design choice, and it is also where the evidence points. burnout research treats burnout as mostly a feature of the job, not the person (maslach and leiter 2016). a review of trials in doctors found that changes aimed at the organisation worked better than changes aimed at the individual (panagioti et al. 2017, jama internal medicine). we say "tend to work better" because that is as far as the evidence goes. the practical reading: change a condition, not a person.

what that looks like:

• do pick one condition and say so. "workload came up, so for the next two weeks nothing new starts until something finishes." small and visible beats big and vague.
• do tell the team what you saw and what you'll try. a survey nobody sees the result of teaches people that their answers are decoration.
• do treat a first reading as a first reading. one check-in is a snapshot. an emerging pattern is worth watching, not deciding on.
• don't try to work out who said it. if you catch yourself guessing, that is the survey telling you the team does not yet believe it is anonymous. say more, guess less.
• don't ask people to explain their answer. it undoes the thing you built.
• don't put the result in anyone's review, including yours. the moment it counts as a score, it stops being honest.

reading team-health signals goes deeper on acting, and on doing nothing, which is sometimes right. and one honest caution: if a layoff or a restructure is under way, the answers will mostly measure fear. wait.

how to start

1. tell your team first. what you will ask, what you will see, what you cannot see, and the one limit above. five minutes at a team meeting does more than any setting.
2. invite everyone on the team. more answers is a better signal, and with a floor of five you need at least five of your reports (not counting you). that is also why we price per team and not per person: it should never be a reason to leave someone out.
3. keep the rhythm. every two weeks, same three questions. treat the first read as a baseline. a trend needs a few check-ins before it means anything.
4. make one change, then say so. that is the whole loop.

from $4 a person, for any team big enough to be anonymous. the starter plan is from $20 a month per team billed annually ($25 monthly), growth from $79 ($99 monthly) for up to ten teams, flat, with 30 days free and no card; the pricing page has the arithmetic and says plainly that flat is not always the lower number. how we work it out names the studies behind everything above and lists what we have not validated. if you are weighing other tools, we have written honest comparisons with officevibe, lattice and the others, each starting with what that tool does well.