Leadership Skills Assessment: How to Build an Evidence-Based Development Plan

Start with the decision the assessment must support

Before choosing a questionnaire, complete this sentence: after this assessment, we need to decide what to develop, for whom, and through which kind of practice. A useful answer might be to identify whether new managers need help with expectation-setting or with following through on delegation. A less useful answer is to discover who our strongest leaders are. The second question invites a broad ranking without explaining what evidence would justify it or what anyone should do next.

This guide covers developmental assessment: finding strengths, learning needs and appropriate support. It does not provide a validated test, a diagnosis of personality, or a basis for hiring, promotion, pay or disciplinary decisions. If the intended use changes to a consequential employment decision, stop and obtain appropriate HR, assessment and governance expertise rather than reusing the exercise under a different label.

Write a one-page brief naming the audience, work situation, behaviours, methods, owners, information boundaries and development decision. Add what the process will not decide. For a regional cohort, identify the shared leadership responsibility and the local differences that matter. A country manager working through distributors and a headquarters manager with direct reports may need comparable judgement without having identical opportunities to demonstrate it.

• Output: an assessment brief with a named development decision and explicit non-goals.
• Do not begin if nobody owns the time, support or practice opportunities that follow.

1. Translate leadership language into observable work

Choose a small set of situations rather than an exhaustive competency dictionary. For example: clarify a delegated outcome, surface a cross-functional dependency, explain a difficult trade-off, or respond to a team member who challenges a decision. Ask what someone would actually hear, see or receive if the behaviour were effective. Replace strong communicator with states the decision, checks understanding and records the next commitment. Replace strategic with compares options against the business objective and makes assumptions visible.

In Singapore, the Institute for Adult Learning's Centre for Workplace and Learning Innovation describes tailored workplace learning that includes on-the-job training, structured mentorship and workplace learning diagnostics. This provides local context for connecting diagnosis with work; it does not validate this guide's assessment design. Use your organisation's existing leadership framework where appropriate, then make the relevant behaviours specific enough to observe. Do not introduce a second competency model simply because it looks more comprehensive.

Separate the skill from the outcome and the conditions. A manager can explain a trade-off well even when the final decision is unpopular. Another may hit a target because demand improved, not because judgement improved. Record the task, expected behaviour and relevant context as three different fields. This keeps evidence about leadership from becoming a retrospective story about business results.

Sources: IAL: Centre for Workplace and Learning Innovation

2. Match the method to the evidence you need

A self-reflection asks what the leader believes about their practice. A stakeholder conversation asks what another person has experienced. A work sample shows a particular response under particular conditions. None is interchangeable with the others. Start with the least burdensome combination that can answer the development question, then add a method only when it closes an important evidence gap.

If the issue is a pattern colleagues experience over time, well-run multi-source feedback may help. If the question is whether a manager can structure a difficult conversation, a short simulation and subsequent workplace observation may be more informative than a broad survey. If the problem is decision clarity, review a suitably redacted decision brief and discuss how it was prepared. Avoid collecting sensitive documents when a de-identified extract or structured description would serve the same purpose.

The US Office of Personnel Management describes work samples as tasks mirroring job activity and notes the role of trained assessors in observing behaviour. Its guidance is about selection, not proof that a home-built leadership exercise is valid. The transferable design idea is task relevance: an exercise should create a fair chance to demonstrate the behaviour being discussed. It should not test hidden knowledge, acting talent or familiarity with one headquarters culture.

Use the method comparison below to choose, not to declare a universal best assessment. A short process with a clear question can be more useful than a comprehensive report that nobody can interpret or act on.

Sources: OPM: Work Samples and Simulations

3. Create behavioural anchors, including not observed

Write the observation guide before anyone takes part. For each behaviour, describe what an incomplete demonstration, a workable demonstration and an adaptive demonstration look like. Keep those labels tied to the exercise, not to the person's identity. A workable delegation conversation, for instance, names the outcome, boundaries, support and review point. An adaptive demonstration also checks the employee's understanding and adjusts support in response to a relevant constraint.

Include not observed as a separate category. It means the evidence does not support a judgement, perhaps because the task never created the opportunity, the observer missed part of the exchange, or the behaviour belongs to a different situation. It must not silently become the lowest score. Also distinguish observation from interpretation: wrote a deadline but did not confirm understanding is evidence; lacks ownership is an interpretation that needs testing.

Do not add the labels into a total leadership score. The useful output is a behaviour profile with examples, uncertainty and context. A person may demonstrate clear direction while needing support to invite challenge. Averaging those findings obscures the actual development decision. The SEED observation card below is a practical design aid, not a psychometric instrument, benchmark or validated scale.

• Situation: a manager delegates a time-sensitive client proposal to a capable colleague.
• Behaviour: clarify the outcome and check understanding without prescribing every step.
• Evidence note: record the words, work product or decision that supports the interpretation.

4. Set information boundaries before inviting feedback

Explain who will see raw responses, who receives the individual summary, what may be shared with the line manager, and how development themes will inform programme design. State how long information is retained and who handles questions or corrections under organisational policy. Resolve these choices before collection, not after a senior sponsor asks to see the spreadsheet. This is an operational design requirement, not legal advice about a particular jurisdiction.

CCL's guidance on organisational 360 initiatives emphasises purpose, trust and clarity about confidentiality and rater anonymity. Its stated model gives the participant ownership of individual results and allows organisational group profiles. That is CCL's approach, not a universal rule or a claim about SEED's service arrangements. Decide and communicate the actual arrangement for your process, and honour it.

Never promise anonymity that a small team, distinctive incident or verbatim comment makes impossible. Follow an established instrument's reporting rules where one is used. In a small cohort, consider confidential facilitated conversations rather than an apparently anonymous report. Do not expose subgroup findings that identify individuals by inference. Provide a route to raise serious conduct or safety concerns outside the development exercise; such matters should not disappear into an aggregated capability theme.

Sources: CCL: Implementing an Organisational 360 Feedback Initiative

5. Prepare observers and pilot the exercise

Ask observers to try the rubric on the same practice response independently, then compare the evidence behind their interpretations. Discuss where the wording is ambiguous, what counts as a demonstration and what the exercise cannot reveal. The purpose is not to force identical opinions. It is to establish a sufficiently shared standard and expose disagreement before it affects participants. When observers disagree, return to the behaviour and evidence rather than letting the most senior person settle the score.

Pilot with people who understand the work but are not part of the first assessment group. Check whether instructions are clear, time is realistic, information is sufficient and the scenario genuinely permits the target behaviour. If you want to observe consultation, the exercise must leave something open to consultation. A script in which every choice is fixed cannot fairly reveal whether someone involves others.

For Singapore and regional Asian teams, distinguish language fluency from the leadership skill unless language proficiency is genuinely part of the task. Offer appropriate preparation, accessible materials and ways to request support. Review unfamiliar jargon and market-specific assumptions. Keep the behavioural standard consistent while adapting irrelevant barriers. Document any meaningful difference in task conditions so a later interpretation does not confuse unequal opportunities with unequal capability.

• Pilot output: revised instructions, agreed examples, clear observer roles and known limitations.
• Stop rule: if the task cannot reveal the intended behaviour, redesign it before using the result.

6. Collect enough evidence without creating surveillance

Create an evidence record with five fields: situation and date, behaviour expected, observation or work sample, relevant context, and confidence in the interpretation. Ask for recent, specific examples where feasible, while recognising that infrequent leadership tasks may need a longer window. Record a strength as carefully as a development need. A process that only collects failure evidence becomes a defect hunt.

Use more than one relevant perspective or occasion where the development question justifies it, but do not set an arbitrary number as a scientific threshold. A stakeholder who rarely sees the work may add less information than a colleague who regularly shares the relevant decision. Let contributors indicate insufficient exposure rather than making them rate every item. Avoid both hand-picking only friendly reviewers and treating a hostile relationship as automatically accurate.

Keep collection proportionate. Do not record meetings, harvest private messages or inspect personal activity simply because technology makes it possible. Use authorised work evidence, normal observation and clearly explained feedback channels. If AI assists with summarisation, use only approved systems and appropriately de-identified material. A human must check the output against the evidence; the model should not infer motives, diagnose personality or produce an employee ranking.

7. Interpret differences instead of averaging them away

When self-reflection, colleague feedback and a simulation disagree, investigate the difference. The leader may know the approach in a workshop but struggle when their own deadlines are at risk. Colleagues may see behaviour that the manager's boss rarely sees. A low rating may also reflect unclear wording, limited exposure or an unresolved working relationship. Disagreement is a question to explore, not proof that one source is dishonest.

Compare like with like: the behaviour, task, time period and conditions should be visible. Ask which finding is supported by examples, where evidence is thin, and what alternative explanation remains plausible. Do not average a self-confidence rating with an observed work sample as though they measure the same thing on the same scale. If the conclusion remains uncertain, plan a better observation rather than disguising uncertainty with a decimal.

Separate capability from opportunity and system constraints. Can the manager explain and demonstrate the behaviour in an appropriate task? Do they have authority, information, time and support to apply it? What happens when they try? A leader who understands delegation but must seek approval for every minor choice may need a governance change as well as practice. Produce a short interpretation: what appears strong, what needs development, what context must change and what remains unknown.

8. Debrief the evidence and agree one development experiment

Give the participant time to understand the findings. Begin with the purpose, methods and limits, then ask what they recognise and what they believe needs correction. Discuss specific examples without asking them to guess anonymous contributors. Separate disagreement about facts from discomfort about impact. Correct factual errors and retain a transparent record of unresolved interpretations; the debrief is not a contest to make the person accept every comment.

Choose one behaviour that matters, recurs often enough to practise and sits within the participant's influence. Agree the next work situation, support, observer and review point. For delegation, the experiment might be to use a clearer outcome-and-boundary brief on the next two appropriate assignments, then ask the colleague to explain their understanding before work begins. Review clarity and unnecessary clarification, not merely whether the participant used the template.

Smither, London and Reilly's 2005 meta-analysis of 24 longitudinal studies found that improvements in multi-source ratings were generally small. This is a reason to avoid equating a feedback report with development. SEED's practical recommendation is to make the follow-through explicit: a bounded experiment, support and evidence for the next conversation. The assessment itself is the beginning of that work, not proof that leadership has improved.

Sources: Smither, London & Reilly (2005): Multi-source Feedback Review

Worked example: assess delegation without ranking managers

This is an illustrative fictional composite, not a SEED client case or claimed result. A regional services business wants managers to reduce avoidable escalations. HR initially proposes a general leadership survey. Discovery instead identifies a narrower question: can managers delegate a meaningful outcome while making risk boundaries and support clear? The first cohort includes managers whose teams have different approval limits across markets.

The assessment uses a short self-reflection, a standardised delegation scenario and one redacted example from normal work. Observers use the same behavioural anchors. One participant performs well in the scenario, but the real brief does not explain when the employee may decide independently. A follow-up conversation reveals that the participant is unsure whether the local approval policy permits delegation. The conclusion is not weak leader. It is evidence of capability in practice conditions, an unresolved authority boundary and an application gap worth investigating.

The sponsor clarifies the actual decision boundary; the participant then tests a revised delegation brief on suitable work. A colleague reviews whether the outcome and limits are understood, while the manager checks the quality of escalations rather than trying to eliminate them. The debrief records what changed, what remains uncertain and whether coaching or programme practice would help. No score is used to rank the cohort, and no improvement percentage is claimed.

Run a proportionate first cycle, then improve the process

A small pilot can begin with one leadership task, a limited cohort and an existing development window. As an illustrative operating sequence, use the first week to agree the brief and information boundaries, the second to pilot the task and prepare observers, the third to gather evidence, and the fourth to debrief and launch the practice experiment. This is a planning example, not a validated timetable. Extend it when feedback availability, accessibility, specialist instruments or workload require more time.

Review the process as well as the participants. Did the task reveal the behaviours? Could observers explain their judgements? Were some groups given fewer opportunities? Did the debrief produce a realistic action? Were information boundaries honoured? Record defects and correct them before expanding to another market or cohort. Scaling an ambiguous rubric multiplies uncertainty rather than creating a stronger baseline.

At the development review, examine new work evidence alongside the original finding. Use a comparable task or situation where possible, and note changes in context. A higher rating may reflect easier work, a different observer or greater familiarity with the exercise. Report what the evidence supports: a behaviour demonstrated more consistently, a remaining support need, or insufficient opportunity to judge. Do not claim business impact or causal programme effectiveness from a pre/post score alone.

Choose an assessment method by the development question

The options below involve trade-offs, not a ranking of validity. Match the method to the behaviour, available expertise and information boundaries. A locally created exercise needs piloting and transparent limitations.

Common mistakes

• Starting with a tool before defining the development decision.
• Rating personality labels, presentation polish or popularity instead of behaviour.
• Treating not observed as a low rating or forcing contributors to judge work they rarely see.
• Adding unlike evidence into a single leadership score with false precision.
• Promising anonymity that small groups or recognisable incidents cannot support.
• Applying identical scenarios across markets while ignoring irrelevant language or context barriers.
• Assuming a feedback report, a higher score or a completed course proves workplace improvement.
• Reusing a developmental exercise for pay, promotion or selection without appropriate governance.

Practical checklist

Frequently asked questions