What a 360° feedback report should include – and how to read it

What a 360° report is for

A 360° report is not a scorecard. Its job is to condense dozens of individual observations into something a leader can work with, without any single observation being traceable to a person. Bracken, Rose and Church (2016) define 360° feedback as a process that collects, quantifies and reports co-worker observations in a way that allows meaningful comparison – between rater groups and over time – and supports sustained behaviour change. Every building block below serves one of those two comparisons. A report that prints only an overall average has quantified something, but it has not enabled a comparison, and it will not change anything.

Definition360° feedback report

The document – interactive or PDF – that brings together a leader’s self-rating and the ratings of supervisors, peers and direct reports on a set of leadership behaviours, shows where those perspectives agree and where they diverge, and – when it is well designed – frames the differences as questions for a development conversation rather than as verdicts.

The seven building blocks of a good report

Rater groups, not one average

The most valuable information in a 360° is disagreement. When direct reports rate delegation two points lower than the supervisor does, that is not noise to be averaged away – it is the finding. A report that collapses all raters into one bar per dimension has thrown away the reason for running a multi-rater assessment in the first place. The price of keeping groups separate is a minimum group size so that nobody becomes identifiable – which is why the breakdown and the confidentiality rules have to be designed together.

Norms that say what they are

A percentile is only as good as the group it is computed against. “Better than 72% of leaders” means little if the reader cannot see who those leaders were, how many there were and when the data was collected. At the start of any instrument’s life the norm is an expert judgement, not a dataset. A report that hides this behind a phrase like “industry benchmark” is not lying with the number – it is lying with the label.

What to distrust

• Decimals. With two or three raters in a group, a difference of a few tenths on a six-point scale sits deep inside the noise. Reliability studies consistently find that more raters are needed than most programmes collect: Greguras and Robie (1998) found little agreement within a rater source; Hensel et al. (2010) needed roughly ten raters for a reliability of 0.7; the two or three peers common in practice gave low reliability. Read tenths as a direction, not a distance.
• Rankings. A league table of leaders turns a development tool into an appraisal, changes how raters answer next time, and rests on a precision the data does not have.
• Unlabelled norms. “Top quartile” with no n, no population and no date is decoration.
• Groups too small to show – shown anyway. If the threshold is two, a group of one must not appear anywhere: not as a bar, not as a comment tag, and not as a residual that can be reverse-calculated from the totals.
• A single item or a single comment as the headline. One low item inside an otherwise strong dimension is a question, not a conclusion.
• Machine-written interpretation. Automatically generated narrative reads with an authority it has not earned. Where free-text comments are summarised or clustered by software, the report should say so – and in a development context, interpretation belongs to a human.

On how many raters it takes, Nowack and Mashihi (2012) summarise the guidance as at least four supervisors, eight peers and nine direct reports for reliability of .70 or higher – and note that two or fewer responses in a group may be inadequate for reliable measurement. That is why a group should either be large enough for its own line or be visibly folded into another.

The debrief: what first, what never

The research on feedback is uncomfortable for anyone who believes a report speaks for itself. Kluger and DeNisi’s (1996) meta-analysis of 607 effect sizes found that feedback interventions improve performance on average – but that more than a third of them made performance worse, and that effectiveness drops as feedback shifts attention from the task to the self. Smither, London and Reilly (2005) found that rating improvements after multisource feedback are generally small and depend on what recipients do next: discussing the feedback, working with a coach, setting goals.

The report, in other words, is the input; the conversation is the intervention. That is the case for a facilitated debrief, whether by an external consultant or a trained internal sparring partner – the trade-offs are set out in consultant-led or self-service?.

1. Context first. How many raters responded per group, which groups were folded, which norm is in use. Without this, every later number is misread.
2. Agreement before gaps. Start with what all groups see the same way – strengths first, then shared development areas. That builds the credibility the harder part needs.
3. Gaps as hypotheses. Take the largest self–other gaps one rater group at a time and ask what the leader would have expected that group to see. Direction before size.
4. Patterns, then items. Only once the dimension-level picture is understood is it worth opening individual items or comments – to illustrate, not to indict.
5. One or two commitments. Close with what the leader will try, with whom, and when the next data point comes – not with a list of twelve improvement areas.

And what never: never open with the lowest score; never read a comment aloud and ask who might have written it; never compare the leader with a colleague’s report; never let anyone leave with the numbers alone and no next step.

How the twelve LEADBeyond 360° views map onto this

The anatomy above is product-independent. For readers evaluating LEADBeyond 360° specifically, this table shows how its twelve report views correspond to it – described, not sold.

Two properties of the report itself count as much as its contents. Every report is reviewed by a consultant before release, who can add a report-level note and per-dimension annotations; both appear in the dashboard and in the PDF. And a released report is immutable: late responses or corrections create a new version that the leader only sees after an explicit re-release, and the norm in use is frozen into each version – a later benchmark update never silently changes a report someone has already read. The PDF (A4, roughly 12 to 16 pages) is rendered from the same charts as the interactive views.