How to select 360° raters: who should give feedback – and who decides?

Why rater selection decides the quality of a 360°

A 360° is not an opinion poll; it is a structured collection of observations. Its value therefore depends less on the questionnaire than on who fills it in. Someone who knows the leader only from meetings cannot judge delegation, coaching or behaviour under pressure – and will tick "cannot assess" or guess. Bracken, Rose and Church (2016) list, among the conditions for effective 360° feedback, formally segmented rater groups, ratings of observable behaviour and enough qualified raters for reliable scores.

"Enough" means more than many programmes plan for. Greguras and Robie (1998), studying 360° ratings of 153 managers, found little agreement within the same rater group; much of the variance came down to individual raters, so more raters are needed than are typically used. Hensel and colleagues (2010) showed that the two or three peer raters common in HR practice produce low reliability. How many raters does a 360° need? covers the numbers; this page is about which people.

DefinitionRaters (feedback givers)

The people who rate a leader's behaviour in a 360° from their own working relationship – grouped as supervisors, peers and direct reports because each group sees different situations and results are reported per group. The leader's self-assessment is the fourth perspective.

Who chooses: the leader, their manager, HR – or all of them?

Four models are common. None is right by default; the table shows what each delivers and where it breaks.

The "picking friends" concern

The most common objection to self-nomination: leaders will choose the people who like them. The concern is legitimate, but manageable. First, friendly colleagues still rate observable behaviour, and a questionnaire that asks, for behaviour items, how often something happens ("never" to "almost always") leaves less room for sympathy than an overall grade. Second, the more people a group contains, the less a single lenient voice weighs. Third, you can prescribe the mix: at least one person per group with whom working together was demanding. To be candid: the studies we cite here examine rater numbers and reliability – not whether self-selected raters rate more leniently. We therefore treat the question as a design decision, not a research finding.

Our recommendation is guided nomination: the leader proposes against the criteria below; HR or the consultant reviews not names but coverage – is a group missing, are critical voices missing, is anyone on the list who hardly sees the leader at work? That review is a conversation inside the nomination window, not a veto.

Selection criteria that hold up

These criteria apply to every group. They are rules of thumb from programme practice, not measurements – but they sort a candidate list reliably.

• At least six months of working together. Anyone with less has rarely seen the leader in more than one kind of situation. Exception: very intensive project work with daily contact.
• Regular contact. Weekly collaboration is a good yardstick; a quarterly meeting is not.
• Observation in the situations that matter. Delegation is seen by those who receive delegated work; behaviour under pressure by those who were there for escalations, deadlines or client crises.
• A mix of critical and supportive voices. Both are data. At least one person per group should have experienced the collaboration as demanding.
• No acute conflict. Someone currently in an escalated dispute with the leader observes from an exceptional situation – that voice can matter, but it should not be the only critical one.
• Willingness and time. 10 to 15 minutes is not much – but it is not available in a closing week or on holiday. Raters can decline; better not to schedule the invitation into such a week at all.

What matters per rater group

Two special cases come up constantly. A single supervisor: in LEADBeyond 360°, their ratings are by default folded into the pooled "all others" view because the anonymity threshold (default: two completed responses per group) is not met; a separate, labelled supervisor column appears only when the consulting team explicitly enables it for that assessment – and the supervisor should know beforehand. Few direct reports: a leader with only two direct reports invites both and compensates with a correspondingly broader peer group, so that at least two groups safely reach the threshold – a report requires that. How the threshold works: Is 360° feedback anonymous?.

Cross-functional and client-facing raters

In matrix and project organisations the formal line is often not the most important working relationship. Assign by the actual relationship: whoever assigns and signs off work belongs with the supervisors, even without people responsibility; whoever is led counts as a direct report, even without a reporting line. Fix that rule before nomination so that every leader in a cycle is treated alike. Client-facing behaviour – the Client & Stakeholder Leadership dimension, for instance – is observed above all by partners and peers on the same engagement; that perspective is in the 360° without surveying clients.

External raters: when clients belong in – and when not

Three reasons usually argue against client feedback in a development 360°: clients almost never see the leader leading their own team; the invitation itself affects the relationship; and the data of external people falls under different agreements than employee data – a question for your legal advisers, which GDPR and 360° feedback puts in context. Clients make sense when the programme explicitly aims to develop client leadership and the relationship is long enough that a structured request for feedback lands as appreciation. LEADBeyond 360° works with supervisors, peers and direct reports; there is no separate client group. If client feedback is essential to your programme, it belongs in the programme design before the cycle is created.

What raters should know beforehand

Raters who understand the purpose and know what is protected answer more honestly and more completely. Three things belong in every invitation – ideally announced by the leader personally before the system invitation arrives:

1. Purpose. Development, not appraisal, promotion or pay – said clearly and kept. Smither, London and Reilly (2005) describe that improvement after multisource feedback is more likely when recipients perceive a need to change, react positively and believe change is feasible – our conclusion: a process experienced as control undermines exactly those conditions.
2. What is protected – and what is not. Individual answers are never shown; results appear as group values once enough people in a group have completed. Free text appears without names, in random order and without AI analysis. But: the leader sees whom they nominated and who has completed – not who answered what. HR sees progress, not results (Can HR see results?).
3. Time and process. 10 to 15 minutes, can be paused, on mobile or desktop; anyone rating several leaders receives one consolidated invitation. Reminders are automatic but can be switched off per cycle. Anyone who cannot rate a person declines with one click – and "cannot assess" is a legitimate answer on any item, excluded from scoring.

What must be in place before the invitation goes out

• Nomination mode and assignment rule (who counts as supervisor, peer, direct report) are fixed before launch and apply to every leader in the cycle.
• The leader has the criteria in writing: six months, regular contact, relevant situations, critical and supportive voices.
• Every group sits well above the minimum; at least two groups reach the anonymity threshold even after declines.
• Direct reports are invited completely; HR or the consulting team has checked the list for gaps.
• The leader has informed nominees personally: purpose, protection, time required, deadline.
• The response window avoids closing periods, holidays and ongoing reorganisations.
• Raters know they may decline, that "cannot assess" is allowed, and who sees the report – see Confidentiality and data protection.

Selecting raters is the part of a 360° that is hardest to automate – and the one that most decides whether the leader believes the result. How a 360° assessment works describes the full process; for partnerships without a classic line, see 360° feedback for consulting and professional-services firms.