Checklist: what to check before you choose a 360° feedback provider

For HR, talent and leadership-development leads and for managing partners introducing a leadership 360° or switching provider. The list is deliberately open – every item can be put to any provider, including us – and says in one line why each item matters.

1. Is the purpose settled before you choose a tool?

• Is the purpose written down – development or appraisal? Once results feed promotion or pay decisions, raters answer differently; hybrids without an explicit rule cost trust.
• Is it settled who owns the report and who sees it? Any combination of leader, consulting team, HR and supervisor works – but fix it beforehand and tell everyone; what HR can see belongs in that rule.
• Is there backing at leadership level? Without a visible sponsor the programme reads as an HR exercise; response rates and seriousness drop.
• Is the communication to leaders and raters prepared? People who do not know what the data is for answer cautiously or not at all.
• Is a repeat cycle planned? A one-off 360° measures a state; only a second cycle shows change. The provider must be able to compare cycles.

2. Who gives feedback – and how many?

• Are the rater groups defined? Supervisors, peers and direct reports are the standard, in project organisations often plus clients. The provider must report groups separately.
• How many raters per group is the target, and how many the minimum? With two or fewer respondents the sample may be inadequate for reliable measurement (Nowack & Mashihi 2012) – see How many raters does a 360° need?.
• Who nominates – the leader, HR or both? Pure self-selection drifts toward sympathetic peers; pure selection by others undermines acceptance.
• What happens with a single supervisor? Shown separately, merged with others, or shown only with explicit agreement – ask for the rule.
• Has rater burden been checked? In small firms the same people rate many leaders; one consolidated invitation, autosave and an option to decline are more than conveniences.

3. How robust is the anonymity you are promised?

• How many responses before a group is shown separately – system rule or setting? Vendor documentation ranges from "no minimum reporting size" to configurable thresholds. Ask to be shown the page.
• What happens below the threshold? Discarded, merged, or shown anyway? Merging keeps the feedback without exposing people.
• Can an individual answer be worked out from group values? If one person makes up the remainder, their answer is a subtraction. Ask whether the system detects that.
• How are free-text comments protected? Unattributed, in random order, hidden when too few have responded. Writing style identifies people faster than numbers.
• What does the leader see about participation? Usually who was nominated and who has completed – legitimate, if said in advance. "Full anonymity" is a promise no process keeps.
• Who at HR or at the vendor can see raw data? Ask about roles, not intentions: technically blocked or merely forbidden? More: Is 360° feedback anonymous?.

The table shows only what vendors document, not what is negotiable; a lower threshold is not automatically better. Broader comparison: 360° feedback software compared.

4. Are data protection and the works council settled before signature?

• Is the legal basis named and documented? Legitimate interest with a balancing test (Art. 6(1)(f) GDPR), a works agreement or, in Germany, § 26 BDSG – not on its own since the CJEU's C-34/21 judgment (2023). Consent alone is shaky at work: per the Article 29 Working Party, employees can give free consent only in exceptional circumstances.
• Is there an Art. 28 GDPR processing agreement – with sub-processors and their locations? "EU hosting" says nothing about where email dispatch, error logs or support access run.
• Is the data protection impact assessment settled? Whether one is required depends on nature and scope (Art. 35 GDPR, the supervisory authorities' lists); the processor must assist you (Art. 28(3)(f) GDPR).
• Are the information duties prepared? Leaders and raters must be told, at collection, the purpose, the legal basis and, where relied on, the legitimate interest (Art. 13 GDPR).
• Is the works council involved early? In Germany, technical systems intended to monitor behaviour or performance are subject to co-determination (§ 87(1) no. 6 BetrVG); personnel questionnaires and appraisal principles need the works council's consent (§ 94 BetrVG). More: 360° feedback and the works council.
• Are free-text comments processed automatically – and has that been reviewed? AI summaries need a named sub-processor, a contract and a traceability assessment. "No, not in the current version" is a legitimate answer.

5. What does the questionnaire measure – and how honest is the methodology?

• Does the questionnaire measure observable behaviour rather than traits? Bracken, Rose and Church (2016) define 360° feedback as a process that captures rater perceptions of specific behaviours and enables comparisons between rater groups and over time. "Is charismatic" is not an observation.
• Is the competency model disclosed – and does it fit how your people lead? A model built for corporate line management rarely fits partners leading project teams without disciplinary authority.
• What does the instrument measure besides behaviour? A behaviour-only 360° answers only what others observe. Whether beliefs, self-efficacy and reactions under pressure are captured too decides whether the report explains or merely describes – see What a behaviour-only 360° misses.
• How long is the questionnaire, and is there a "cannot assess" option? Long rater surveys cost response rate – ask for the estimated time per rater group; forced answers without a basis for observation produce noise and must stay out of the calculation.
• What is the benchmark based on, and does the report say so? An expert norm is legitimate if labelled as one; an "industry benchmark" with no sample behind it is marketing.
• What validation evidence exists – and how honest is it? A correlation between two scales of the same questionnaire, completed by the same raters, is not evidence of business impact. Ask about sample, method and criterion measure.

6. What happens after the report?

• Does the report show groups separately, self against others and individual items – or only totals? One average across all raters throws away the point of the format: the differences between perspectives – see What is in a 360° report.
• Does a person review every report before the leader sees it? A report with three responses in a group, one plainly careless, should not go out automatically.
• Who runs the debrief, and is it part of the offer? Smither, London and Reilly (2005) found, across 24 longitudinal studies, only small average improvements after multisource feedback; change, they argue, depends on recipients acting – discussing it, working with a coach. Software alone does not do that – see Consultant-led or self-service?.
• Is there a process after the debrief? Development goals, a conversation with the supervisor, a check-in after three months. Without follow-through a 360° is information, not feedback.
• Does a released report stay unchanged? Late responses and corrections should create a new, labelled version – not silently alter the report the conversation was built on.

7. What is in the price, and how do you run the vendor conversation?

• What is in the price – licence, set-up, debriefs, repeat cycles? A low price per leader is usually a software price; facilitation is added or absent. Compare total cost per cycle.
• Which model fits the population? For hundreds of leaders without a conversation, a survey tool is cheaper; for twenty partners whose results shape careers and culture, self-service rarely suffices.
• In which languages are questionnaire, report and interface available – completely? "Supports German" can mean only the buttons are translated and the items must be entered yourself.
• What happens to the data after the programme? Retention, export, comparability with a later cycle – in writing.
• What does the product not do? A vendor that names no limits has not thought about it or is busy selling. Ours: no self-service for mass roll-outs, no AI processing of free text, not a selection instrument.

How to use the list in a vendor conversation: pick the ten items that decide your programme – usually purpose, minimum group size, free text, legal basis, hosting, debrief and what the price covers. Ask each question open-ended, without revealing the answer you hope for, and ask for the documentation page or a demonstration rather than an assurance. "Usually" and "on request" count as open. Two vendors, same questions, same order – then you compare answers, not presentations.