How to add a 360 without turning it into an average

Before designing the form, write down the decision that will use the result. A development conversation, a mobility decision, and a compensation decision involve different risks and evidence requirements.

The number of participants does not fix a poorly framed question. Ten people may repeat the same impression, observe only one part of the work, or use incompatible criteria. The unit being assessed also needs to be defined. A question about cross-functional leadership requires episodes in which the person influenced others without hierarchical authority. Asking people who only receive direct instructions to provide that rating produces opinions about a different behavior.

1. Assign a question to each source

This rule prevents reputation from being treated as evidence. It also avoids asking an internal customer to judge a technical decision whose criteria they could never know.

Confidentiality changes what can be answered. In small groups, promising anonymity may be unrealistic. The process must explain what will remain confidential, who will see comments, and when a serious allegation will follow a separate review path.

2. Preserve provenance before summarizing

Suppose someone receives high ratings within their own team and low ratings from two areas that receive their work. A middle-of-the-range average hides the most important signal: the experience changes at the interface.

The difference has several possible explanations. Expectations may be incompatible, quality may be inconsistent, access to information may be unequal, or the criteria may be ambiguous. The 360 does not automatically decide among them. It locates the differences so the organization can return to specific episodes.

Not every source should have the same weight for every dimension. Weighting should follow access and relevance, not hierarchy or volume. If peers observed coordination for months and the manager only saw the final outcome, the design must preserve that difference.

3. Define a rule for disagreement

Disagreement is not an error by default. Nor is it a hidden truth that the average will resolve.

Classify the discrepancy before acting. It may arise from different facts, different contexts, different interpretations of the scale, or insufficient evidence. Each source calls for a different response.

If two sources observed different periods, extend the period or limit the conclusion. If they applied incompatible criteria, revise the anchors. If adverse episodes are concentrated in one relationship, examine that interface. If no one provides sufficient facts, preserve the uncertainty.

This guidance connects with Not observed, not applicable, and insufficient evidence do not mean the same thing. Those states must survive the process; they cannot become zeros used to complete a matrix.

4. Separate feedback from the decision

Feedback can guide reflection and practice. A high-impact decision also requires validity for that use, known rules, traceability, and a route for review.

Avoid automatically reusing the same result. An interpretation created for development does not become valid for compensation simply because the database already exists. Changing the use changes incentives, candor, consequences, and evidence requirements.

CIPD’s evidence review of performance feedback notes that its effect depends on factors such as credibility, delivery, timing, and the opportunity to participate in the conversation. It also discusses perceptions of fairness when information is accurate, complete, and open to response. CIPD, Performance feedback: an evidence review (opens in a new tab).

5. Return an interpretation, not a verdict

The feedback should separate three outputs: supported points of agreement, differences that need explanation, and areas without sufficient evidence.

For development, each conclusion can be connected to an observable practice and a new opportunity for review. For mobility or compensation, a material discrepancy may block the conclusion until relevant evidence is obtained.

CIPD places performance reviews within a broader process and recognizes that they may support development or decisions such as pay. That range of uses reinforces the need to define the purpose before choosing the instrument. CIPD, Performance reviews (opens in a new tab).

A good 360 does not produce a single point of view. It produces a defensible interpretation of partial perspectives. The report should show who observed what, where sources agree, and what remains open. In that way, disagreement provides information instead of disappearing into a decimal.

To extend this reading, see How to design a calibration session that does not turn into a negotiation over scores, Self-assessment, manager, or peers — what each source contributes, and How to design a scale that reduces differences in interpretation, which develop complementary dimensions of the problem.