ingridleiria Chief of Staff at Series C
Everything that was implicit now has to be made explicit, and every time something becomes explicit it gets slower. An approval chain appears. Legal is involved earlier. Finance wants a business case for a purchase that used to be a message. Each addition is individually defensible and the sum is a company that takes six weeks to do what used to take a day, with nobody able to name the decision that made it so, because there wasn't one.
The second failure at this stage belongs to the role rather than the company. Without a written mandate, the job dissolves into whatever the chief executive did not get to that week. At six hundred people that is a full-time occupation that produces nothing anyone can point to at the end of the year, and it ends with a capable person being quietly moved sideways because nobody could describe their contribution.
This is the inflection point where companies either learn to run at scale or spend three years discovering that what got them here has stopped working. The role narrows here rather than widening: most of the surface it covered at earlier stages now belongs to people who are genuinely better at it.
When to use this, and when not to
Use it between roughly three hundred and a thousand people, when there is a functional leadership team that runs itself, when the role is being scoped around a specific programme such as an integration, a market entry, a pricing change or a platform migration, or when the diagnosis in chief-of-staff-by-stage returned the mandate holder archetype. Use it when the presenting complaint is that everything has become slow and everyone can name a different culprit.
Do not use it while the main value is still reconciling the functional reports and the role has no named programme; that is cos-at-series-b. Do not use it once decisions are routinely made in the chief executive's name under written authority, and the company is preparing for a listing or a sale as a continuous state; that is cos-at-series-d.
The adjacent case: where the mandate is really a programme that should have a full-time programme director and a team, say so. program-management covers running it as a discipline, and taking a large delivery programme personally is how the rest of the role's surface gets abandoned.
What you need before starting
The mandate in the chief executive's own words, and what it is instead of. Missing: write the two or three candidate versions and ask which one they mean. If the answer is all of them, the negotiation about scope is the first piece of work, not a preliminary to it.
The steering group, and specifically who can say no. A programme at this size crosses functions whose leaders can each stop it. Missing: name the three people whose objection would halt the work and get them into the steering group before the first milestone, not after the first collision.
A baseline of whatever is being changed. Cycle time, cost, headcount, revenue, error rate, whichever is the point. Missing: measure it for two weeks before changing anything. A programme with no baseline cannot be defended at month nine, when the questions start.
The statutory and governance calendar. Board and committee dates, audit, statutory filings in every entity, anything with a legal date attached. Missing: build it in week one from whoever holds the corporate calendar. In the United Kingdom and much of the Commonwealth that is a company secretary; elsewhere it is usually the finance lead or general counsel, and in a charity or a public body the governance officer or clerk. Discovering one of these dates late is the fastest way to lose the board's confidence.
How the executive team currently works. Its meeting, its agenda, what it actually decides. Missing: attend three of them and take notes on decisions made rather than topics covered. The ratio is the diagnosis.
How often the current process is bypassed. Exception counts, emergency approvals, purchases split to stay under a threshold. Where a connected purchasing or ticketing system is available, count these directly; where it is not, two conversations with people who deliver work give the answer. Missing: ask what they route around and why. A process with a high bypass rate is not a control, it is a fiction with an administrative cost.
The method
Write the mandate to one or two things and get it agreed in writing. The test that makes it real: what will be observably true in twelve months that is not true now. A mandate that cannot pass that test is a theme, and themes cannot be resourced, defended or finished.
Set the programme spine before doing any of the work. Named milestones with dates, one owner per workstream who is not you, a steering rhythm, and a written statement of what is out of scope. The out-of-scope list is what stops a twelve-month programme becoming an eighteen-month one, and it must be written at the start, when refusing is cheap.
Baseline before changing anything. Two weeks of measurement, published. This costs a fortnight and buys the ability to prove the programme worked, which at this size is the difference between a mandate renewed and a mandate quietly wound down.
Sequence by irreversibility, not by tractability. Do the parts that are hardest to undo, and the parts where people can leave, before the parts that are merely fiddly. Systems work is tractable and feels like progress, which is exactly why it seduces programmes away from the risks that actually decide the outcome.
Put every proposed new process through the admission test. Three questions and two requirements, below. This is the mechanism by which professionalisation happens without the company slowing to a stop, and applying it consistently is worth more than any single process you design.
Give every new process a sunset date. Twelve or eighteen months, at which it is re-argued or it lapses. Processes never remove themselves, and a company at this size accumulates them at a rate no annual review can absorb.
Build the governance calendar backwards from the fixed dates. From each statutory or board date, work back to when material must be drafted, reviewed, and approved. Publish it a year ahead. Nothing on this calendar should ever be discovered late, and the calendar is how you make that a property of the system rather than of your memory.
Raise executive team effectiveness, because nobody else will. Whether the team functions as a team, whether the right things reach it, whether its meetings decide anything. Do it with observations rather than adjectives: how many decisions were made in the last four meetings, how many were revisited, how many agenda items were reports rather than decisions.
Keep and defend a short list of decisions that stay fast. Write down which decisions are deliberately low-approval and who holds them, and defend the list by name when someone proposes a review step. A company that has made everything careful has made nothing important.
Plan the mandate's ending from the start. Every mandate resolves into a permanent owner, a closed programme, or a business. Name which, and the date you expect it, in the first month. Programmes without a planned ending become departments.
The process admission test
Every proposed process answers three questions before it is adopted, in writing, in four lines each.
What failure is this preventing, and has it actually happened here. Not could happen: has happened, with the instance named. Most proposed process at this stage is imported from somebody's previous company, where the failure did happen.
Who does it slow down, and by how much. In days per instance, times instances per quarter. This converts a vague sense of caution into a number that can be weighed against the failure it prevents.
What is the exception path, and who can grant it. A process with no exception path becomes a reason to route around the whole system, which is worse than the risk it was preventing.
Plus two requirements: a named owner, and a sunset date. A process with neither is a habit, and habits are what make a company slow in ways nobody can trace.
Where the answers do not justify the cost, the honest response is no, and the role is senior enough at this stage to say it. Saying it consistently is most of what protects the company's speed.
Governance, when it becomes real
Board meetings become formal, committees appear, and the company acquires obligations about what is recorded and when. Directors join who were not there at the beginning and who read the pack rather than living the company. Minutes become a document that matters, and some decisions must legally be made by the board rather than by the chief executive.
board-and-investor-management carries the mechanics. What changes at this stage is the standard: material goes out far enough ahead to be read, the same numbers appear in the pack as in the operating review, and a restatement is explained rather than made silently. Independent directors form their view of management's reliability from exactly these details.
Complexity arrives alongside it, none of it interesting until it is expensive: multiple legal entities, employment law in places nobody has worked before, data protection obligations that differ by jurisdiction, transfer pricing. Keep one register of these obligations with an owner against each.
Decide alone, or escalate
Decide alone: everything inside the mandate, the governance calendar, the shape of board material, the planning process, what reaches the chief executive, and whether a proposed process passes the admission test.
Escalate: executive hires and exits, anything with material legal or financial exposure, anything that changes the strategy, and any judgement about a member of the leadership team.
The judgement specific to this stage is when a founding leader has outgrown their role. Everybody can see it before anybody says it, and it is the most damaging thing at this size to leave alone. The observation goes to the chief executive, in private, once, as behaviour and consequence rather than character, and then it is theirs.
Worked example
Situation. Kesterbridge, a field service software company. Six hundred and twenty people across four countries, eleven months after a 130 million Series C, revenue 74 million. The company had acquired a 140-person competitor for 55 million, closing in March, and the chief executive gave the Chief of Staff a single mandate: integrate it inside twelve months. The acquired company had a different product on a different technology stack, an overlapping customer base of about ninety accounts, and a founder-chief executive who had agreed to stay for eighteen months.
Task. Deliver an integration where the twelve-month test was written as: one sales organisation with one price book, no customer holding two contracts on incompatible terms, the acquired team's senior people still present, and the combined engineering organisation on one release cadence.
Action. The mandate was written and agreed in the first fortnight with an explicit out-of-scope list: no platform migration, no rebrand, no changes to the acquired company's engineering practices in year one. That list was disputed twice and held both times.
The wrong turn ran ten weeks. The programme started with systems, because systems are tractable: one customer relationship system, one human resources platform, one finance close. Three workstreams, real milestones, visible progress. In week eleven three of the acquired company's five senior engineers resigned inside a fortnight, and the exit conversations said the same thing three ways: nobody had told them what their job was now, and the integration they could see was about tooling.
The sequencing rule was rewritten as sequence by irreversibility. People leaving is irreversible; a customer relationship system running twice for a year is merely annoying. The systems workstreams were slowed deliberately and two things moved to the front: each of the acquired company's forty-one senior and mid-level people got a named role, a manager and a written scope inside six weeks, and the ninety overlapping accounts were assigned a single owner each with a rule for which contract governed.
The admission test earned its place in month five. Finance proposed a business case requirement for any purchase above 10,000 across the merged company. Question one: the failure had happened once, in the acquired entity, at 240,000. Question two: it would touch roughly 380 purchases a quarter and add about six days each, which is 2,280 days a quarter across the company. Question three: no exception path was proposed. The version adopted set the threshold at 75,000, added an exception granted by any executive, a named owner, and a sunset review at eighteen months.
Result. At twelve months three of the four written tests passed. One sales organisation and one price book landed in month nine; every overlapping account had a single owner and a governing contract by month seven; of the forty-one people scoped in the first wave, thirty-six were still there at the anniversary.
The fourth test, one release cadence, was missed and was declared missed rather than redefined. It landed at month fourteen, when the programme closed into a permanent owner, the engineering leader, with two workstreams formally dropped as not worth their cost: the human resources platform consolidation and a shared support queue.
The ten weeks lost to systems-first sequencing cost more than the delay suggests. Three senior engineers is roughly a year of hiring in that market, and the exit conversations made clear the resignations were preventable by a conversation that cost nothing.
A second scenario, where it goes differently
A four-hundred-person company where the offered mandate was to help the chief executive professionalise the company. It failed the twelve-month test immediately: nothing observable would be different, so nothing could be finished, and two attempts to get the mandate narrowed produced enthusiasm and no decision.
The rule applied was to pick the narrowest defensible version, write it, act on it for ninety days, and come back with evidence rather than a request. The version chosen was decision speed for routine spending, because it was measurable and universally complained about.
The baseline took two weeks: a standard purchase between 5,000 and 50,000 took a median of 31 days from request to approval, across four approvers, with 19 percent of purchases split into smaller amounts to stay under a threshold, which is the bypass rate that tells you a control is fictional. After ninety days, with two approval layers removed, one threshold raised, and an exception path written, the median was 9 days and the splitting had fallen to 4 percent.
That number, not the argument, got the mandate written. It was agreed in month four as two things: decision rights and the operating cadence, with the second half of the year scoped around a specific pricing change.
What changed the method was not the company's size or stage. It was that the chief executive could not narrow the mandate in the abstract and could choose easily between concrete alternatives once one of them had a measured result attached. Where a mandate will not be written, do the narrowest version and let the evidence write it.
Output
MANDATE
Mandate: [one or two things, in one sentence each]
Twelve-month test: [what will be observably true that is not true now]
Explicitly not: [the out-of-scope list, agreed at the start]
Steering group: [names, including everyone who could say no]
Ends by: [date] into [permanent owner / closed / a business]
Agreed with the chief executive on: [date]
PROGRAMME SPINE
| Workstream | Owner (not you) | Milestone | Date | Irreversible? | State |
BASELINE
| Measure | Before | Target | Now | Measured how |
PROCESS ADMISSION TEST [one per proposed process]
Prevents: [failure, and the instance where it happened here]
Slows: [who, days per instance, instances per quarter]
Exception: [path, and who grants it]
Owner: [name] Sunsets: [date]
Verdict: [adopted / adopted with changes / declined]
DECISIONS DELIBERATELY KEPT FAST
| Decision | Who holds it | Why it stays fast |
Failure modes
No written mandate. Recognise it when you cannot say in one sentence what will be true in twelve months. The role dissolves into leftover work within a quarter. Fix by writing the narrowest defensible version and acting on it.
A mandate that is a theme. Professionalise, align, improve. Recognise it because there is no test that could fail. Fix by converting it to one observable outcome with a date.
Sequencing by tractability. Recognise it when the visible progress is all systems and tooling while the people and customer questions are unanswered. Fix with the irreversibility rule and re-sequence.
Owning the workstreams yourself. Recognise it when your name is against delivery rather than against the spine. At this size that guarantees the rest of the role is abandoned; see program-management for staffing it properly.
Process admitted without the test. Recognise it retrospectively by the bypass rate: when people split purchases or route around an approval, the process failed its exception question. Fix by re-running the test on the process that is being bypassed rather than enforcing it harder.
No baseline. Recognise it at month nine when someone asks whether the programme worked and the answer is a narrative. Two weeks of measurement at the start prevents this permanently.
Governance discovered late. Recognise it as a scramble before a board or filing date. Fix by building the calendar backwards from fixed dates and publishing it a year ahead.
Silence about a leader who has outgrown the job. Recognise it when several people have said it to you privately and nobody has said it upward. The role's proximity is exactly what makes this its job.
Edge cases
Two mandates that conflict. A cost programme and a growth programme given to the same person will resolve into whichever is asked about most often. Name the conflict in writing, ask which wins under pressure, and record the answer.
The mandate is finished early. Close it formally, hand it to a permanent owner, and go back for the next one rather than extending scope to stay busy. Extending a finished programme is how a role loses its edge.
The programme is failing and it is your programme. Report it at the steering group before it is asked about, with the recovery option and the cost of each. A late admission at this size does more damage to the role than the failure itself.
An acquisition arrives mid-mandate. Integration will consume the year whatever the current mandate says. Renegotiate rather than absorbing both, and get the displaced mandate reassigned or paused in writing.
Multiple entities and jurisdictions. Keep one register of obligations with an owner against each. Where specialist advice is needed and the budget is not there, write down the obligation and the fact that it is unadvised, and put it in front of the board rather than carrying the risk quietly.
Quality bar
- The mandate is written, agreed, limited to one or two things, and carries a twelve-month test that could fail.
- The out-of-scope list was written at the start and has survived at least one challenge.
- Every workstream has an owner who is not the Chief of Staff.
- A baseline was measured before anything changed, and the current number sits beside it.
- Every process adopted this year passed the admission test, and has a named owner and a sunset date.
- The governance calendar is published a year ahead and nothing on it was discovered late.
- The list of deliberately fast decisions exists and has been defended by name at least once.
- The mandate has a planned ending, with a named permanent owner.
Adapting this to your context
The defaults come from venture-funded companies of three hundred to a thousand people carrying an integration or a market entry. The label is shorthand for a shape.
- "Series C" as a label. It means three or more layers with managers of managers, a leadership team that runs without the chief executive, statutory obligations with real dates, and approvals multiplying faster than anyone chose. A hospital group, a multi-site retailer, a mid-sized professional firm or a family business in its third generation is here whatever it has raised.
- The admission test numbers. A 75,000 threshold and 380 purchases a quarter came from one company. Compute your own from the failure's actual cost against days lost per instance times instances per quarter. The arithmetic is the method; the figures are not.
- The twelve-month test. Fits a mandate inside an annual cycle. Where the cycle is an academic year, an electoral term or a three-year grant, set the test at that horizon and keep it observable.
- The governance calendar. Built from board, audit and filing dates. Substitute regulator inspections, funder reporting, accreditation visits or licence renewals, whichever carry legal dates for you.
- What not to change. A mandate limited to one or two things with a twelve-month test that could fail, and a baseline measured before anything changes.
Related skills
cos-at-series-b precedes this, and the synthesis habit built there is what makes a mandate holder credible when they report on their own programme. cos-at-series-d follows, when decisions are routinely made in the chief executive's name under written authority.
program-management is the discipline for running the mandate itself, and the place to go when it needs a team rather than a person. board-and-investor-management and board-deck carry the governance mechanics. strategic-plan-and-action-plan produces the strategy the mandate usually implements. annual-planning-and-headcount is the cycle the mandate has to survive. decision-memo is the format for every escalation. process-documentation-sop writes up whatever passes the admission test.