How it works
Ask the people who were there. Then have the conversation.
Every survey Facets runs — a team rating itself, a team describing its leader, a whole organization answering at once — follows the same rules: questions about what people actually saw, nothing reported below a floor, nobody ever quoted, and honesty about how much the numbers can hold.
Four products, one set of rules
The team survey
Everyone on a team rates everyone, including themselves, and the results become a 60-minute debrief. Sold per run, per team. It is the product the rest of this page walks through, because it is the one the other surveys borrow their machinery from.
The Leadership 360
The same privacy stance pointed upward: a team describes what it is like to work for one leader, and the leader gets a brief that compares what they expected with what was said. Sold per leader.
Engagement cycles
Everyone in an organization answers the same questions once, and the answers come back as group numbers with error bars — for the organization, each unit, and each manager. Sold per cycle, for the whole organization.
The manager tools
The deliberate exception: not a survey but a record a manager keeps — goals, check-ins, written assessments. It is a performance record, it says so on every screen, and it is kept structurally apart from the surveys above. Sold per manager.
The team survey: what people are actually asked
First, a short calibration
Three to five minutes before any rating: the definition of each behavior with worked examples at the low, middle and high end, one practice judgment per behavior with the keyed answer, and a reminder of the three ways people usually get this wrong — rating someone you like higher on everything, giving everyone the same score, and rating potential instead of the last three months.
This is called frame-of-reference training, and it is the single cheapest thing you can do to make ratings comparable between people.
Then, behavior — not personality
Every peer item has the same stem: in the last three months, how often did this person… followed by something you could have watched happen. Five frequency options from “rarely or never” to “almost always”, plus not enough opportunity to observe — which is treated as missing data, never quietly scored as a middle value.
Five core dimensions, two items each: dependable contribution, helping and backup, open communication and voice, constructive disagreement, and keeping the team on track.
3 further dimensions — raising standards, inclusive participation, respect under pressure — are switched on per team when you set it up. When they are on, everyone rates all 6 of their items, except on a team of exactly four, where they are split so each item still reaches two raters without asking any one person to rate everything. On a team of three both raters see them all, because with only two raters per person there is nothing to split.
On a team of five that is somewhere around 52 to 76 individual judgments — roughly fifteen minutes, because anchored frequency questions are fast to answer.
Two written answers about each teammate
After the frequency items for a given person, two short free-text questions: what they should keep doing, and what would help. These are the only place in the survey you write prose about a named colleague.
Your name is dropped from them at scoring time, before anything is stored for analysis. They are never quoted back — not in the team pack, not in anyone’s individual note — and an automated check rejects any pack that reproduces a recognizable run of words from one.
Three questions about working with them
On a seven-point agreement scale rather than a frequency scale, because they ask something different: whether you would want this person on a project like this one, whether you go to them, and whether you can rely on them. They feed the two indices in your results; the reliance question is also pooled across every pair of teammates into one team-level trust figure, and nothing else.
Then the team, the network, and yourself
34 questions on team climate — psychological safety, direction and role clarity, coordination, task cohesion, relationship conflict, information flow, team efficacy, reflexivity, viability, meetings and decisions, and respect and fairness — answered about the team rather than about individuals. Up to 6 more are asked only of teams that work apart or that depend on other teams.
A few short “who do you go to” questions that produce a picture of how work and influence actually flow, as opposed to the org chart.
Then your own ratings of yourself, and — this is the part most tools skip — your prediction of how your teammates rated you. The gap between what you predicted and what they said is usually the most useful number in the whole report.
The self module also asks 13 questions about how being on this team is going for you: whether you feel safe raising a concern, whether your workload is sustainable, whether you care about the work, whether you are learning, whether you belong, whether you are making progress on work that matters, and whether you expect to still be on this team in a year. That last one is a retention signal, so it is worth knowing you are being asked. All 13 are reported only as team averages, with a count of how many people answered below the midpoint, and only once 4 people have answered; no individual answer to any of them is shown to anyone.
What the scoring does, and what it refuses to do
Raters differ in how generous they are, and that difference is large — a substantial majority of the variance in multisource ratings is about the rater rather than the person being rated. So scores are also reported perceiver-centered: each rater’s own average is subtracted before comparison, which removes most of that.
Every individual score carries a reliability band that depends on how many people rated you, because a mean of two ratings and a mean of six are not the same measurement. Climate scores are only reported as a team score when members actually agree enough to justify treating them as one; when they do not, the report says the team is divided instead of averaging the disagreement away.
Below three raters, nothing is shown.
On a three-person team, individual peer scores are withheld entirely.
Three reports, three audiences
The team pack — everyone sees it
Climate scores with agreement and spread, the network picture, and a 60-minute meeting agenda built from what your results actually say. No individual is named or ranked.Your own note — only you see it
What your teammates see you doing well, one thing to work on, how your self-view compared with theirs, and how confident the whole thing is given how many people rated you. Never sent to your manager, and not readable by your team lead, your workspace owner, or anyone else with an account. The one exception is a facilitator your team has named by email — and only that person.The facilitator view — if you have named one
Data-quality signals, dyads whose ratings of each other are unusual, and anything the automated checks flagged for a human to look at before release. Notes that trip a flag are held rather than published — and when a facilitator is attached, nothing at all reaches a person about themselves until that facilitator approves it. This view is readable only by them; if no facilitator is named, nobody can open it.
The debrief is the intervention
The agenda opens with what the team is already doing well, moves to two or three things worth changing, and ends with commitments specific enough that you could tell in six weeks whether they happened.
The written packs are constrained on purpose. No trait adjectives, no ranking language, no speculation about why someone behaves as they do, and no causal claim about your results — “teams higher on coordination tend to do better in published research” is allowed; “this is why your project slipped” is not.
Two of those constraints are checks rather than instructions. Every score-like number in a pack is matched against the scored data before release, and any six-word run shared with somebody’s written answer is treated as a leak and rejected — with the survey’s own dimension names excluded, since those are our words and not anyone else’s. A pack that fails either check does not go out; the deterministic version goes out instead.
Six weeks later, a two-minute pulse
The debrief ends with commitments, and the lead records them with an owner and a check date. Six weeks on, everyone gets a two-minute pulse: for each commitment, how far has it happened, on a scale from “not at all” to “fully”; then the same two questions the run asked about feeling able to speak up and about workload, so the two figures sit side by side.
No peer ratings, deliberately: six weeks apart, ratings of colleagues measure how the raters recalibrated, not what changed. The pulse closes on the day the team set, the results go to everyone at the same moment, and nothing is shown until three people have answered. It is part of the run, not an extra.
The same questions, at organization scale
An engagement cycle asks everyone in the organization the same climate statements the team survey uses, about the team each person actually works in — so a unit’s numbers mean the same thing whether they came from a cycle or a team run — plus questions about their own work. Results come back as group numbers with error bars, for the organization, each unit, and each manager, with groups below the reporting floor left out rather than quietly included.
What a cycle deliberately does not produce is this page’s debrief pack: no agenda, no individual notes. A cycle tells you where to look; a team that wants to work on itself runs a team survey.
Leadership feedback, separately
The same privacy stance applied to upward feedback for one leader: 3 to 25 raters, 14 behavior items across seven dimensions, and a brief that compares what the leader thought their team would say with what the team said. Nothing is reported below three responses, comments are grouped into themes and never quoted, and a facilitator can be given the brief first and decide whether it is released.
It is a different product from the team survey on this page, sold separately, and it has its own page.