For leaders

What you write about them. What they say about you.

Two tools for the same job, and they never touch. The manager tools are your record of the people who report to you — goals, quarterly check-ins, 1:1 agendas, assessments. The Leadership 360 runs the other way: your team describing what it is actually like to work for you, under a promise about who reads it that you cannot override.

Get startedSee a write-up

The manager tools are $20 a month for one manager, and the people you track never need an account. The Leadership 360 is $99 for one leader, bought once per review.

The manager tools

The write-up you owe them, done properly.

Goals, quarterly check-ins, 1:1 agendas and end-of-year assessments, for the people who report to you. Built on one assumption: the useful part is your own judgment, and everything else exists to make you write it down honestly. Three decisions follow from that, and they are not in the other tools.

You write first. Then you read what colleagues said.

The assessment page runs in a fixed order: your answers, then colleague input, then the draft. A manager who reads three colleagues’ views and then writes has produced a summary of the colleagues. A manager who writes first has an account of their own, which the colleagues can then confirm or contradict. Only the second is worth having, and the only thing that reliably produces it is doing them in that order — so the order is enforced rather than suggested.

It asks what you owe them.

“What did you say you would do for them that you have not done?” is in the quarterly set and the annual set. It lands in the document under What you owe them, which is what makes an assessment a two-way record rather than a verdict. It is also the question a manager is most likely to skip, which is why it is asked outright instead of left to occur to you.

It will not assign the rating.

Below, meeting or exceeding is set by you, in one place, next to a restated definition of what the bands mean — and the document fails its checks if you set one without writing why. It is the most consequential line in the write-up and the only defensible author of it is the person who will have to say it out loud.

Beside them is the idea board: one question to the whole team, answers that carry names or not on terms you set before anybody posts, and a loop that is not finished until every idea has a status and a sentence back from you.

Buying for every manager at once, rather than for yourself? The Organization plan is these same tools with consolidated billing, priced per manager.

For organizations

The year

What you actually do with it.

  1. Add someone

    A name is enough to start — a half-filled record is more useful than one nobody made. Role, scope, strengths, what they need from you, and what you are watching can come later.

  2. Set goals, and sharpen them

    Goals per year, with a measure and a target. Any goal can be rewritten on request into something observable, because “be more proactive with stakeholders” cannot be assessed and everybody knows it at the point of assessing it.

  3. Check in each quarter

    Progress against each goal, a band, and what changed. Quarter boundaries are fixed, and a check-in is never requested for a quarter that has not finished.

  4. Prepare the conversation

    A guide written against the failure mode of the band you recorded — vagueness when someone is below, blandness when they are meeting, over-promising when they are exceeding. Plus 1:1 agendas that fit the time you actually have, and follow-up actions pulled out of your notes afterwards.

  5. Ask colleagues, if it will help

    Up to ten people, three written questions, no rating scale. A five-point average from two colleagues reads as data and is noise — it would be the single most misleading number this could show you.

  6. Write the assessment

    Quarterly or end-of-year. Four guided questions or seven, each with the reason it is being asked underneath. Then a draft, your edits, your band, your rationale, and an export written to be handed to the person it is about.
Two worked examples, whole

Guardrails

What the draft is not allowed to do.

These are checks that run over the text before you see it, not instructions in a prompt. A rule that only exists in a prompt is a hope.

Never quote a colleague

Any run of eight or more words shared with someone’s raw answer is caught, and the draft reaches you with that failure named at the top of it rather than quietly — and it is checked against what they actually wrote, not against the anonymized copy, because checking the stripped version would pass by definition. With three respondents, a recognizable clause identifies its author to a manager who read the answers an hour earlier.

Never attribute a view to someone who was not asked

A colleague noun next to a reporting verb is flagged. Your own account of how someone works with others is fine and always was — that is your observation, and you are allowed to have made it with nobody’s help.

Never claim a pattern from one person

One reply is used, and used as one person’s view. Two people agreeing is the smallest honest claim of a pattern, and the draft, the prompt and your own panel all say so.

Never open a formal process

Discipline language is rejected outright. An assessment records; it does not initiate. Where the record genuinely warrants escalation the app says so and points at the people whose job that is — and does not script the step for you.

Never state a number that is not in the record

Every figure in the prose is matched against the underlying data before you see it.

What this is not

Four things the manager tools do not do, that you should know before you start.

There is no view for the person being assessed.

No acknowledgement, no dispute, no version history. That is a real feature and a bigger one, and half of it would be worse than none — so the export is the delivery mechanism, and it is written to be read by them rather than about them.

Only you can see your records.

There is no HR role and no organization-wide view on a single manager plan. That is deliberate, but it has a cost: if you leave, the record is unreachable. Export is available on every person and every assessment for exactly that reason.

Colleague input is named, not anonymous.

Respondents to staff feedback requests are told, before they type, that you will read it with their name on it.

Generating a draft sends the record to a third party.

Snapshots go to Anthropic’s API. It is said at the button that triggers it rather than buried in a policy page, and the snapshot carries no email address — the data structure has no field for one.

The wall

Everything above is a performance record. Everything below is not.

Facets is two products behind one login, and this page is where they meet without touching. Everything above — goals, check-ins, bands, assessments — is a performance record: you writing down what you think of someone who reports to you. Everything below is the opposite kind of data. The team survey and the Leadership 360 are developmental: people answer under a promise that their individual ratings reach the person they are about and nobody else, and the whole thing rests on that promise holding.

Both being in one product creates exactly one serious risk, and it is not technical: a colleague somewhere decides the developmental promise is a formality, because the same company and the same nav bar also run a performance record. Every structural decision here exists to make that conclusion wrong, and to make it visibly wrong to somebody who goes looking.

Exactly one table joins the two, and it is the record of a summary somebody chose to send you. No other table on either side references the other, and no policy lets even that one be read back through into the review, its raters or its responses. There is no join to write, so nobody writes one by accident. That share is the only route between the two — and attaching it to their record is something you do by hand, because inferring that link would be the first automatic connection between the two and that is the one thing the design exists to prevent.

So the rest of this page is the other half, and nothing on it can reach the half above it.

Who can see what

The Leadership 360

The other direction: what your team says about you.

You invite your team — the people who report to you, and peers if you want them — and they describe what it is actually like to work for you. 14 behavior items across 7 dimensions, 2 overall questions and 4 written prompts, answered through a link. Your raters never sign in, and between 3 and 25 of them can be asked.

You say what you think they will say.

You answer the same 14 items twice: once as your own view of how you lead, and once as your guess at how your team would answer. Most instruments compare you to your team and stop there. The gap that actually changes behavior is the second one — between what you expected to hear and what you heard. A leader who is surprised has learned something. A leader who predicted it exactly already knew, and has a different problem to solve.

Nothing is ever quoted.

Written answers become paraphrased themes. In a group this small a direct quote names its author, and a distinctive clause does it almost as well. So this is a check rather than an intention: no six-word run from anyone’s answer may appear in the brief, and one that does is caught before the brief reaches you.

There is a floor, and below it you get nothing.

Under 3 responses there are no scores, no themes, and no written answer leaves the database at all — not a partial view, nothing. Item-by-item detail needs 5. A theme only one person raised is folded into the pattern it belongs to, or dropped. The floor costs you a report in exactly the case where a report would have identified somebody.

The brief

What comes back, and what it may not say.

Two strengths. At most two things worth working on, each with something concrete to try. A section on where your own view and your team’s diverged. The comment themes. A plan for taking the results back to the team, which is the part most 360s leave you to invent. And a statement of how much confidence the whole thing deserves, which names how many people answered.

Behavior, never character

Task and behavior language only; trait words are rejected. No ranking, and no score against other leaders — you are not managing other leaders, and a percentile answers a question nobody asked.

Every number matched against the evidence

The same rule as the assessment drafts above: a figure in the prose that is not in the scored data does not ship. If the model fails its schema twice, a deterministic writer produces the same brief from the same inputs with no model involved — the same sections, not a degraded version — and the record says which one wrote it.

Follow-up questions, asked the way a coach would

The comments are where most of the value is, and most of them arrive thin. So after a rater has written theirs, a model asks one or two short questions built from that person’s own answers — what it looks like in practice, an example of when it happened, what came of it. It is shown that one rater and nothing else: not another rater, not your self-ratings, not your context note, not the confidential box. It may not ask them to name a colleague. They can skip it, and you can switch the step off for a review.

A facilitator can hold it back

Name one and they read it first — with the routing flags and the confidential channel — and decide whether it is released to you. Without one it releases automatically when the checks pass and nothing is flagged; when something is flagged it is held and you are told to add a facilitator. Every access is logged. Afterwards you can record commitments against the brief and re-run it later against them.

Anonymity here is an effort, not a guarantee.

Themes are combined across everyone who answered, and a specific example is described only when more than one person raised it — a single person’s example is folded into the pattern it illustrates. That last rule is an instruction to the writer rather than a check, because nothing mechanical can tell an incident from a pattern. Raters are told this in these words, on the page where they consent, before they type.

$99 for one leader, one review a year. Included in the Practitioner plan. This is not the same offer as the manager tools above, which is why the two are priced separately on the pricing page rather than bundled into one number.

Get startedWhat it costs

Questions

Is this the same thing as the team survey?
Neither half of this page is, and none of the three ever join. The team survey is anonymous peer feedback that goes to the person it is about; nobody with an account can read an individual's ratings, including you. The manager tools are your own records about people who report to you. The Leadership 360 is a third thing again — feedback about you, developmental, readable by you and by a facilitator if you attach one. Exactly one table joins any two of them: the record of a summary somebody chose to send you, and no policy lets even that be read back through into the review or its raters.
Can the person read what I wrote?
You can export the assessment as a document and give it to them, and it is written to be read that way — fairness notes first, the rationale beside the band, and a closing line saying what fed it. There is no in-app view for them yet. Separately, in most places a written characterization of an employee is personal data they can ask to see, so the honest assumption is that they will read it. The app says so wherever you type about someone.
Does the AI decide the rating?
No. The band is set by a human, in one place, and the document fails its checks if there is no rationale written beside it. The draft can be generated; the judgment cannot. Everything a generator writes is also checked before you see it: no colleague is quoted, no view is attributed to someone who was not asked, no number appears that is not in the record, and no language that opens a formal process.
Why is the colleague input named rather than anonymous?
Because with two or three respondents anonymity is a fiction. A manager who receives 'one person said you miss deadlines' from a pool of three has effectively been told who. Promising anonymity there and failing to deliver it is worse than never promising it, so the page says plainly, before anyone types, that their manager will read it with their name on it. Names are then stripped before anything reaches a generator, so the document itself cannot name a colleague.
What happens if a colleague does not want to answer?
Declining is a real button with its own confirmation. You learn that they declined so you stop waiting, and nothing about why. Any draft text they had started is discarded — a declined request still holding half an answer would be a record of something somebody chose not to say.
How many people do I need for a 360?
At least 3 have to answer before anything is shown, and at most 25 can be invited. Below 3 you get told there were not enough responses and that is all — no scores, no themes, and no written answer leaves the database, because a "partial view" from two people is exactly the thing the floor exists to prevent. Item-by-item detail needs 5. If you have fewer than 3 people who can honestly speak to how you lead, the instrument is the wrong tool and a conversation is the right one.
Can my raters be identified from the brief?
The work to prevent it is real and it is not a guarantee, and we would rather say so than promise something we cannot enforce. Comments are never quoted — that part is a check, not an intention, and a brief containing a six-word run from anyone's answer is rejected before it reaches you. Themes are combined across everyone who answered, and a specific example is described only when more than one person raised it; a single person's example is folded into the pattern it illustrates. That last rule is an instruction to the writer rather than something a machine can verify, because nothing mechanical can tell an incident from a pattern. Raters are told all of this, in these terms, before they type.
Who sees my leadership brief?
You. A review is bound to the account that created it, so it comes back to you and to nobody else by default. If you name a facilitator, they see a version of it first — with the routing flags and the confidential channel — and decide whether it is released to you. Your own manager does not see it, and there is no organization-wide view. Nothing from a 360 lands in anyone's performance record: if you want to share part of it you do that deliberately, and what arrives is a copy.
What is the confidential box for?
When a facilitator is attached, raters get one field that goes to them and to nobody else — not to you, and never to a model. It exists because the alternative was including destructive-leadership items in the instrument itself, and the research on upward feedback is fairly clear that inviting negative ratings of a leader tends to cost the raters afterwards. A confidential channel to a professional does the same job without putting the rater's name on it in your report.
What does it cost?
The manager tools are $20 a month for one manager, or $16 per manager per month billed annually across an organization — priced for the people writing reviews, not for everyone receiving one. The Leadership 360 is $99 for one leader, bought once per review. Questions about that, or about anything on this page, go to facets.team/contact.
Get startedThe team survey, separately