Questions d'entretien : Chief of Staff

I design the cadence around decisions the leadership team must make, not recurring meetings.

• I interview each executive to map decision types, required inputs, owners, and latest useful decision dates.
• I separate weekly exception reviews from monthly strategy and talent reviews that need deeper preparation.
• Every forum gets a purpose, pre-read deadline, decision owner, and cancellation rule when no decision is needed.

A strong agenda gives most time to decisions and the rest to exceptions needing executive help.

• Routine metrics go in a pre-read; only material variance, its owner, and a proposed response enter the room.
• Each decision item states the question, decision maker, options, recommendation, and time box in advance.
• I close with decisions, owners, due dates, and messages leaders must carry back to their functions.

I test whether the meeting produces decisions or unblocks work at a cost worth paying.

• For four cycles I track preparation, decisions, closed actions, attendance, and executive hours consumed.
• Repeated status updates move async; a forum with the wrong decision makers gets redesigned rather than lengthened.
• I remove duplicate forums only after specifying where their essential decisions and information will move.

I make the review exception-based and require owners to arrive with a decision or request.

• The scorecard arrives 24 hours early, with red and amber metrics linked to short variance notes.
• Agenda time goes to threshold breaches, cross-functional risks, and trade-offs requiring executive authority.
• Items needing no decision, escalation, or resource change stay in the written update.

I use one metric and priority hierarchy across rhythms, with each serving a different time horizon.

• Weekly reviews handle exceptions, blockers, and near-term commitments against quarterly objectives.
• Monthly reviews examine trends, resource shifts, and dependencies that one week cannot reveal.
• Quarterly planning closes the prior cycle, resets priorities, and feeds new owners and milestones into weekly reviews.

I keep a small linked set showing priorities, performance, decisions, dependencies, and commitments.

• The core set is a company scorecard, quarterly priority register, decision log, dependency tracker, and action register.
• Each artifact has one owner, a refresh deadline, and a source of truth such as Notion, Linear, or the finance model.
• Completed cycles are archived so rationale remains traceable without old pages competing with current data.

I translate each strategic choice into a few outcomes showing whether that choice works this cycle.

• Objectives describe the desired change; key results use baselines, targets, owners, and dates rather than project milestones.
• I ask what decision would change if each metric moved or stayed flat.
• Initiatives sit beneath the OKR so teams can change methods without quietly changing the intended outcome.

A good key result is measurable, owned, time-bound, and tied to an outcome the business values.

• I write movement from a known baseline to a target by a date and name the data source.
• I avoid binary launch language unless launch itself removes measurable risk or fulfills a contract.
• The owner must explain controllable levers and the lag between action and metric movement.

I run OKR check-ins as short reviews of confidence, evidence, and intervention needed.

• Owners update actuals and confidence beforehand, then explain only meaningful changes.
• A low-confidence result includes the cause, recovery option, resource request, or recommendation to reset.
• I record the executive decision and next checkpoint so the same risk is not rediscovered weekly.

I separate a failed execution plan from an invalid target before recommending a change.

• I verify data, original assumptions, leading indicators, and whether the owner controls the main levers.
• If the outcome still matters, I present a recovery plan with resources and cost to other priorities.
• If the premise changed, I document the reset and preserve the original score rather than rewriting history.

I force prioritization by limiting company objectives and exposing the resource cost of every addition.

• Each proposed OKR names the strategic choice it supports and work the team will stop.
• I map shared people and budgets across objectives to expose capacity counted twice.
• Lower-priority work can remain on roadmaps, but it gets neither company-OKR status nor executive review time.

I change an OKR only when its strategic premise, measurement quality, or external constraint materially changes.

• A difficult target or weak execution is not enough; those remain visible in the score and recovery plan.
• The owner brings evidence, proposed wording, and downstream impact to the forum that approved the goal.
• The decision log retains the original target, change date, rationale, and approver.

A useful log explains what was decided, why, and when to revisit it without reopening the debate.

• I capture the decision, accountable decision maker, date, options, rationale, and material assumptions.
• I link analysis and separately record affected teams, communication owner, and resulting actions.
• High-uncertainty decisions get a metric or date trigger for review; settled decisions stay closed.

I translate each decision into owned commitments before the operating cycle moves on.

• The decision log stores the call; the action register stores deliverables, owners, dates, and dependencies.
• I send a same-day readout and give owners a short correction window.
• Overdue actions return by exception with the blocker and requested escalation, not a generic reminder.

I distinguish new evidence from discomfort and reopen a decision only for the former.

• The log shows rationale, assumptions, and review triggers, so a challenger must identify what changed.
• New evidence returns to the named decision maker with its impact.
• Without new evidence, I restate the decision and route execution issues to the action owner.

I assign one decision maker based on accountability for the outcome and make other roles explicit.

• I identify who owns the result, who holds essential expertise, and who must execute the choice.
• A RACI or RAPID map helps only if it names one final decider and avoids universal veto power.
• I escalate only choices beyond delegated authority, changes to company strategy, or unresolved material risk.

I provide enough context for execution, not a transcript of the executive discussion.

• The note states what changed, why now, who is affected, required actions, and where questions go.
• Depth varies by audience, but rationale and numbers stay consistent across leaders, managers, and employees.
• Sensitive decisions use a timed cascade so managers are briefed before their teams hear the news.

I structure a QBR around performance against plan, lessons, and decisions for the next quarter.

• The pre-read covers scorecard results, financial variance, customer signals, priorities, and dependencies from one dataset.
• Meeting time focuses on causes of major gaps and trade-offs functional reviews cannot resolve.
• The output records decisions, changed priorities, owners, and required updates to the operating plan.

A QBR is an internal operating review; a board meeting combines governance, strategic counsel, and director accountability.

• The QBR goes deeper into functional causes, owners, and corrective actions for the next quarter.
• Board materials elevate company performance, cash, risk, strategic choices, and matters requiring formal approval.
• I use the QBR to resolve inconsistencies before they appear as conflicting board numbers or narratives.

I work backward from the meeting through data, narrative, review, and distribution stages with named owners.

• Finance and functional metrics get source owners and an early freeze so the CEO reviews stable numbers.
• The CEO reviews the storyline before slide polish; the CFO and General Counsel review financial and sensitive sections.
• I reserve the final buffer for consistency checks, appendix requests, and secure distribution, not first drafts.