How to Build a 360 Feedback Assessment in 2026, Question by Question

You've decided to run a 360. Now you have to build the thing.

Search online for 360 feedback questions and you get a list. Twenty-five questions for a 360 review. The twelve most insightful ones. Forty examples sorted by rater group. Those lists are fine as far as they go. None of them tells you what happens after somebody answers.

That's the part that breaks. A 360 is five or six people answering about one person, and their answers have to come back together as a report that person can actually read. Picking the questions is the easy half. Building them so they add up is the half that decides whether the report says anything.

So this covers the whole build. What to ask, how to word it, what scale sits behind it, how the score comes out, and how you keep one subject's raters attached to that subject.

What a 360 Feedback Assessment Needs Before You Write a Question

Maya manages a team of six. Her director wants her in the leadership programme next quarter, and the programme opens with a 360.

That's the unit of work. One person, one cycle, one report. Everything you build serves it.

An employee engagement survey asks a lot of people how they feel about the company, and you read it in aggregate. A 360 asks a few people about one person, and you read it one person at a time. Run one as if it were the other and you get a report nobody can use, because Maya can't act on a company average and the company can't act on Maya's delegation score.

Two decisions come first. Both change what you write, so make them now instead of editing forty items later. Still choosing where to build it? The 360 feedback tools comparison covers that. This article is about what goes inside.

Pick Four to Six Competencies, Not Twelve

A competency is a behaviour somebody can watch you do. Communication. Delegation. Handling disagreement. Doing what you said you'd do by the day you said you'd do it.

Pick four to six. Each one needs three or four questions behind it before the number means anything. So twelve competencies is forty items, and your raters are people with jobs. They start carefully and click through the back half.

Four competencies at four questions each gives you sixteen rated items and two open ones. Ask anyone to do that and they'll finish it. Ask them for forty and you get eleven half-finished forms and a report built on the first two competencies.

A short assessment everyone completes beats a thorough one half of them abandon.

Decide Who Rates Whom Before You Build

Maya's panel: Maya, her director, four peers, five of her reports. Eleven raters in four groups.

Those groups are the reason a 360 is worth running. A peer sees things a director never will, and a report sees something else again. Splitting the results by group is what turns a pile of ratings into something Maya can act on. Which means the assessment has to know which group each rater is in before it can score anything.

Set the minimum per group before you send a single invite. Three is our recommendation. At two, people can work out who said what, and the moment they suspect that, they stop telling you the truth. If a group can't reach three, fold it into another group or leave it out of the report.

Say the threshold out loud in the invite. Raters who know where the suppression line sits will answer straight, and the ones who have to guess will hedge. Get the mechanics of an anonymous survey setup right before you promise anything.

The Question Types a 360 Feedback Assessment Actually Needs

Three question types do all the work. Everything else is decoration you'll regret when you read the report.

Get the mix right and the report writes itself. Get it wrong and you've got fifty comments and no numbers, or sixteen numbers and no idea what they mean. If you want a shape to start from, there are questionnaire templates worth copying, and any decent survey builder holds all three types in one flow.

Rated Behaviour Statements Carry the Score

These are the bulk of the assessment. A statement about something the subject does, and a scale the rater picks from.

"Maya explains a decision clearly enough that her team can repeat it to someone else."

Anyone on Maya's team has either seen her do that or hasn't. Sixteen statements like that, four per competency, give you a score per competency and a score overall.

Use the same scale on all sixteen. Five points is enough: never, rarely, sometimes, usually, always. Frequency works better than agreement here. "Strongly agree" invites the rater to tell you their opinion of Maya. "Usually" invites them to tell you what they saw last month.

Two Open Questions Carry the Meaning

Numbers tell Maya where she stands. Words tell her what to do on Monday.

Two is the right number. Ask for more and you'll get "n/a" three times.

• What should this person keep doing, and what does it make possible for you?

• What's one thing this person could change that would help your own work most?

Somebody busy can answer both in two sentences. Neither invites a character assessment. The second one is deliberately narrow: one thing, and its effect on the rater's own work. "What are their weaknesses" gets you either nothing or something Maya should never read.

One Question That Says Which Group Answered

The question that makes the whole report work is the boring one at the top.

"What's your working relationship with Maya?" Manager, peer, direct report, self. One choice, required, first page.

Without it you've got sixteen averages and no way to see that Maya's reports rate her delegation at 4.4 while her peers rate it at 2.8. That comparison is the most useful thing in the exercise. You can branch on the answer too, so a direct report gets slightly different wording from a peer while both feed the same competency score.

Ask for it explicitly rather than inferring it from who you sent the link to. People forward links.

Writing the 360 Behaviour Statements, Question by Question

A good statement is specific, observable, and about one thing. Most bad ones fail on the third.

Four Statements That Work, and Why

Take delegation. Four statements, each getting at a different part of it.

1. Hands over work with enough context to start without coming back for more. Observable, and it separates delegating from dumping.

2. Leaves the method to whoever does the work, once the goal is agreed. This catches the manager who delegates the task and keeps the method.

3. Steps in when something is stuck rather than when something is late. Specific enough that a rater has a moment in mind.

4. Hands out work that stretches people a little. Gets at whether delegation is developing them or just clearing a desk.

Read those four and you can already picture the manager who scores high on one and two and low on three. That's what a competency needs: four statements that can disagree with each other. If all four always move together, you've asked the same question four times. Plenty of employee feedback form examples make exactly that mistake.

The Statements to Cut

Four shapes produce noise. Cut them on sight.

• Two behaviours in one statement. "Communicates clearly and listens well." A rater who's seen one and not the other has no honest answer, so they pick the middle and you learn nothing.

• Anything about attitude or personality. "Has a positive attitude." Nobody can watch an attitude, and Maya can't change one on Monday.

• Anything only the manager can see. Ask a direct report to rate strategic planning and you get a guess with a number attached.

• Comparisons. "Is a better communicator than most managers here." Every rater is comparing against a different set of managers.

The test: could the rater name the moment they're thinking of? If they can't, the statement is measuring how they feel about the subject, which is a different exercise from the one everybody agreed to.

Building a 360 Assessment in Involve.me, Question by Question

Sixteen rated statements, two open questions, one relationship question. In involve.me that's one funnel, and you build an assessment as a Score-based Outcomes funnel so the score can route the report.

Pick the funnel type first. You choose it upfront when you start from scratch, it comes preset when you start from a template, and the AI Agent picks it from your prompt. Changing it later is possible and it isn't a five-minute job, so decide now. If every rater should land on the same closing page, a Thank You page is enough. If you want the report to change with the score, you want Score-based Outcomes.

Which Element to Build Each 360 Statement In

Build every rated statement as a "Single/Multiple Choice" question with five answers: Never, Rarely, Sometimes, Usually, Always. That element carries "Individual Score & Calculation", and so do "Image Choice" and "Dropdown". That option is what turns sixteen answers into a competency score.

Two details that save a rebuild. Reach for one of those three elements when the answers need to add up. And the "Score" element has a second mode, "Correctly Answered Questions", which counts correct answers at one point each; that mode is for quizzes, so leave yours on "Individual Score", which sums the values of the selected answers.

My view is that the scale belongs in the question element and nowhere else. A 360 needs sixteen items that all count the same, and every formula you add is a thing you'll have to explain to Maya when she asks how the number was made.

Setting Answer Values So the Scale Means Something

Turn on "Individual Score & Calculation" on the question, then set each answer's value by hand. Never is 1, Rarely 2, Sometimes 3, Usually 4, Always 5.

Do it the same way on all sixteen. One question scored 0 to 4 while the rest run 1 to 5 shifts the total, and nobody finds it later.

Sixteen items on a 1 to 5 scale gives you a range of 16 to 80. That's a fine number for the engine and a useless one for a human. Divide by the item count when you show it, so Maya reads 3.4 out of 5 instead of 54 out of 80. She already has a scale for the first one.

Generating the Item Bank With AI

Writing sixteen statements that all pull their weight is the slow part. Describe the competency to the bulk question generator and it drafts the items. Or ask the AI Agent for the whole funnel and edit down from what it builds.

Then rewrite every statement. AI drafts land on the generic version of a competency, and the whole point of four statements per competency is that they can disagree with each other. Same goes for anything an AI survey generator hands you. Keep the structure, replace the wording with the behaviours that actually matter where you work.

Keeping One Person's 360 Raters Together

Eleven people are about to fill in the same assessment about Maya. Next month another eleven fill it in about somebody else.

A questionnaire has no idea who it's about. That's the problem the question lists never mention, and it has a small answer.

Pass the Subject in the URL With a Hidden Field

In involve.me this is one hidden field and one link per subject.

1. Add a custom hidden field to the funnel and name its parameter subject_id. Parameter names are lowercase, with underscores or dashes instead of spaces.

2. Give it a fallback value you'll spot in your data, like unassigned, so a link somebody stripped the parameter from doesn't come back blank.

3. Send each rater the full funnel URL with the value on the end: https://your-org.involve.me/360-review?subject_id=E-4417

The panel shows a live URL parameter preview at the top, so you can copy the shape instead of typing it from memory. Full hidden field setup guide in the help centre.

Every submission from that link now carries subject_id. The value shows up in your analytics under the name you gave the field, which is how you pull one subject's eleven raters as a group. And because a custom field becomes available to answer piping as soon as it exists, the same value can be written into the questions the rater reads.

Two things that would otherwise cost you a cycle. The short ivlv.me URL doesn't accept parameters, so rater invites need the full funnel URL. And values containing spaces or apostrophes have to be URL encoded first, which is the main reason to pass an ID rather than a name.

Use an employee ID where you can. It groups just as well, it needs no encoding, and it keeps the subject's name out of a link that gets forwarded around. If those IDs already live in your HR system, remote_id does the same job without being created in the editor first: append ?remote_id=E-4417 and the value saves with the submission.

One Funnel Per Cycle, Not Per Rater

Build one funnel and reuse it for everybody.

The relationship question sorts the rater groups, the hidden field sorts the subjects, and the funnel stays a single thing you maintain. Duplicate it per subject and you've got twelve funnels drifting apart by March, each with its own slightly edited statement three, and no way to compare this cycle to the last one.

That matters most when you run cycles for several managers at once, which is the normal case. It's the same argument behind the HR services funnels generally: one build, many cycles.

One funnel that serves every cycle is also the only version you can improve.

Scoring a 360 Feedback Assessment So the Report Says Something

A score on its own tells Maya almost nothing. She reads 3.4 out of 5 on communication and has no idea whether that's what her team thinks or what she thinks.

Setting up calculations and score-based outcomes came up twenty times in our own support inbox over the last 3 months. Reading that list, the pattern I see is people building all the questions first and working out the report afterwards.

Do it the other way round. Decide what the report has to show, then score for that.

The Self-Versus-Others Gap Is the Finding

Maya rates her own delegation at 4.6. Her reports put it at 2.9. That gap is the single most useful thing the assessment produces, and averaging her self-rating in with everyone else's destroys it.

So keep the self score out of the group average and show it beside them. Maya reads her own 4.6 against her director at 4.0, her peers at 3.6 and her team at 2.9. Nobody has to explain what that means to her.

A gap the other way is worth just as much. Somebody who rates themselves at 2.4 while their team says 4.1 is underestimating what they're good at, and that's a different conversation with a different next step.

One thing to settle at build time rather than after. Every rated statement needs a "Not observed" answer, because a peer who has never seen the behaviour will otherwise pick the middle of the scale and quietly move your average. Give it a value of zero so it adds nothing to the sum. It still sits in your item count, so divide by the number of statements actually answered rather than by sixteen when you write the report.

Score Bands and What Each One Should Trigger

Bands turn a number into a sentence. Three is plenty: below 3.0, 3.0 to 4.0, above 4.0.

Decide what each band does before you decide what it's called. A band that changes nothing is decoration.

• Below 3.0 on any competency. That competency goes into the development plan, and it's what the follow-up conversation opens with.

• 3.0 to 4.0. Solid. Mention it in the conversation and leave it alone.

• Above 4.0 with a self score more than a point lower. A strength the subject doesn't know they have. Say it out loud, because this is the band that gets skipped most often.

In involve.me those bands are score tiers on the "Graph" or "Scorecard" element, which renders the number as a gauge, a progress ring or a bar chart in your own colours and labels. Want the closing page itself to change with the band? Use a Score-based Outcomes funnel with multiple outcomes, one page per band.

Show one gauge per competency rather than one for the whole assessment. An overall 360 score flattens the four competencies you spent the whole build separating, and Maya can't do anything with it.

What Happens After the Last 360 Rater Submits

The conversation with Maya is the deliverable. There's usually a two week gap between the last rater submitting and anyone having it.

Build the follow-up as one workflow in involve.me, triggered on "Completed Submission" for the funnel. The first "Send Email" node sits directly on the trigger and thanks the rater, which is the only email a rater ever needs. A "Send Internal Email" node tells you a submission landed, so you can watch the panel fill up without refreshing anything. A "Wait" node holds the rest until the cycle closes.

The report goes to the subject and their manager, and the raters should hear that in the invite. Raters who think their words are going straight to the person write differently, and usually less.

This is where automated email sequences that follow up earn their place: the reminder to raters who haven't answered, the report to Maya once the panel is complete, the nudge to her manager to book the conversation, and the check-in eight weeks later asking what actually changed. All of it runs off the submission data the assessment already collected.

The eight-week check-in is the email everybody skips, and it's the only one in the sequence that tells you whether the 360 was worth running.