Civil Service Judgement Test cheat sheet

Decision rule: gather evidence only until additional information would no longer change the choice; beyond that point, delay costs more than it buys.

Escalate when a decision exceeds your delegated authority, carries legal, security or reputational risk, or commits another team's resources.

Urgent (time-pressured) and important (high consequence) are separate dimensions; rank options by impact multiplied by likelihood, not by who asked loudest.

Reversible decisions should be made quickly and corrected on evidence; irreversible or one-way-door decisions justify full analysis and sign-off.

Every decision needs a recorded audit trail: options considered, evidence used, assumptions made, and the reason the chosen option won.

Consult subject-matter experts for facts, but the accountable owner decides; consultation is not delegation of accountability.

Test each option against five filters: lawful, ethical, affordable, deliverable, and consistent with published policy.

When evidence is incomplete, state the assumption explicitly, choose the least harmful option, and set a review trigger rather than stalling.

The defensibility test: ask whether a reasonable outsider, shown the evidence and rationale, would accept the decision.

Triage grid: urgent and important = do now; important but not urgent = schedule; urgent but not important = delegate or minimise; neither = drop.

A task that unblocks other people's work outranks an isolated task of the same size, because its delay propagates.

Critical path is the longest chain of dependent tasks; slippage there moves the whole delivery date, while slack elsewhere absorbs delay.

Report slippage as soon as it is foreseeable, with a revised date and a recovery option; late bad news costs more trust than the slip itself.

Parkinson's law: work expands to fill the time available, so explicit shorter deadlines usually compress delivery without reducing quality.

Roughly 80 per cent of the outcome comes from a small share of tasks; identify the minimum viable deliverable that still meets the need.

Silent over-commitment is the main cause of missed deadlines; scope must be renegotiated explicitly rather than absorbed quietly.

Maintain a single prioritised list; parallel lists cause duplicated work and dropped items.

Batch small similar tasks together and protect one uninterrupted block for work requiring sustained concentration.

Feedback targets observable behaviour and its impact, not personality; character labels trigger defensiveness and are hard to act on.

Peer disagreement is handled first in a private, fact-based conversation; escalation is a later step, not the opening move.

Escalate when an issue repeats despite direct attempts, threatens delivery or safety, or involves bullying, harassment or discrimination.

Separate intent from effect: most colleagues intend well, but the effect still needs correcting and both facts belong in the conversation.

Restate the other person's position until they agree it is accurate, then respond; this lowers defensiveness and tests your understanding.

Give credit publicly and criticism privately; both are more effective and lower the cost of the message.

Make specific asks (task, owner, deadline) rather than general appeals for help; vague requests are the commonest reason help never arrives.

If a colleague is overloaded, negotiate priorities with their manager rather than reassigning their work yourself.

On joining a team, map the informal decision routes as well as the org chart; formal authority rarely describes how work actually moves.

Bottom line up front: state the conclusion or the request first, then context, then detail, because readers frequently stop early.

Match channel to stakes: complex, sensitive or contested messages need a live conversation; simple factual exchanges suit written channels.

Influence depends on framing benefits in the audience's terms, supported by evidence from a source that audience already trusts.

Use plain English and define acronyms on first use; unexplained jargon signals exclusion and reduces compliance.

Follow a verbal agreement with a short written record of who will do what by when; this prevents later disputes about what was decided.

Bad news sequencing: headline, cause, impact, remedy, next step - never bury the problem under background.

Summarise what you heard and check it before responding; a clarifying question is cheaper than acting on a wrong assumption.

Stakeholder grid by power and interest: high power with high interest is managed closely; low power with low interest needs only monitoring.

Confirm understanding by asking the recipient to restate the action, not by asking whether they understood.

Separate the stated want from the underlying need; customers usually arrive with a solution in mind and the need is the problem to solve.

Publish service standards - what will be done and by when - and measure performance against them rather than against impression.

Before refusing a request, check whether policy actually forbids it; many refusals are habit rather than rule.

Apply policy consistently to comparable cases and record any discretion used, because an unexplained exception becomes the new expectation.

When you cannot help, explain why and signpost an alternative route; refusal with reasons preserves trust, unexplained refusal does not.

Treat recurring complaints as a signal of process failure and fix the cause rather than each instance.

Share only the personal data the other party needs for the purpose; surplus detail is a data protection risk with no service benefit.

Acknowledge contact within a stated timeframe even when the substantive answer will take longer.

Collect feedback at the point of service, while the experience is fresh, rather than weeks afterwards.

Change curve sequence: shock, resistance, exploration, commitment; resistance is a normal stage, not evidence of disloyalty.

Communicate change by answering five questions: why, what, when, what it means for me, and what stays the same.

Pilot a change at small scale, measure the result, and scale only what the evidence supports.

Keep the outcome fixed and vary the method; altering the goal mid-delivery requires explicit agreement from the owner.

Resilience technique: separate the event from your interpretation of it, because the interpretation drives the stress response and can be revised.

Sustained pace requires planned recovery periods; they are part of performance, not a reward for finishing early.

After a setback, run a short no-blame review of what happened, why, and what changes; blame suppresses accurate reporting.

Consult frontline staff early when designing a change, because they see the failure modes first.

Watch for sunk-cost reasoning: past investment is never a reason to continue, only expected future benefit is.