ingridleiria Chief of Staff at Series B
There is now a real leadership team, and that changes the job completely. Every functional leader can see their own area clearly and none of them can see the company. The chief executive receives five confident reports every month, each internally consistent, and the truth sits in the space between them where nobody is looking.
The specific failure this prevents is a company that discovers a contradiction between two functions in the last month of a quarter, when it is expensive and nobody can be blamed for it, because each function was doing what its own numbers said. Sales forecasts on a pipeline the roadmap cannot support. Finance plans headcount against a hiring rate recruiting has never achieved. Customer success reports healthy retention on a definition that excludes the accounts that churned. Each report is honest. Together they describe a company that does not exist.
This is also the stage where the role either becomes strategic or becomes an expensive coordinator, and the difference is whether you produce a view nobody else in the company can produce. Running a good meeting is no longer a differentiator at this size; there are fifteen people who can do that.
When to use this, and when not to
Use it when a functional leadership team owns its own numbers, headcount is roughly between a hundred and three hundred, the board has become a governance body rather than a supportive investor call, or when the diagnosis in chief-of-staff-by-stage returned the synthesist archetype. Use it when the complaint is that the numbers do not agree, or that the chief executive keeps being surprised late.
Do not use it while the role still owns functions and is handing them over; that is cos-at-series-a, and synthesis attempted on top of functional ownership produces neither. Do not use it once the role carries a named mandate such as a market entry, an integration or a listing, and the value has moved from reconciliation to programme execution; that is cos-at-series-c.
The adjacent case: where the functional reports are genuinely incomplete rather than contradictory, the work is reporting hygiene before synthesis, and weekly-status-update and revenue-forecast cover the inputs.
What you need before starting
The monthly report from every function, in whatever shape they arrive. Missing: ask for what they already send their own teams rather than commissioning a new report. A synthesis that starts by adding reporting work to five leaders begins with five people against it.
The plan of record, with its assumptions written down. Not the target: the assumptions underneath it, such as sales cycle length, hiring rate, conversion, and churn. Missing: reconstruct them from the last planning cycle, circulate them, and let the leaders correct them. The corrections are themselves a finding.
A metric dictionary. One definition per term, agreed by the functions that use it. Where a connected analytics or reporting tool is available, take the definitions from the queries that produce the numbers rather than from what people say they measure, since the two differ more often than not. Missing: build it before the first synthesis. Reconciling numbers defined differently produces confident nonsense and destroys the document's credibility in one month.
The last two board packs and the questions asked in the meetings. They show what the board is watching and where the chief executive is exposed. Missing: ask the chief executive which three questions they most dislike being asked.
Standing access two layers below the leadership team. Skip levels, or attendance at team meetings. Missing: arrange it explicitly with the leaders rather than informally, because informal access reads as surveillance and will be treated as such.
What the chief executive does with the document. Whether it feeds a decision, a board conversation or their own thinking. Missing: ask directly after the second one, and change the format to fit the answer.
The method
Fix the inputs before attempting synthesis. One report shape, the same fields, the same dates, the same definitions. Not a new report: a template applied to what leaders already produce. Until the inputs are comparable, the synthesis is an opinion.
Read every report against the plan first, not against each other. For each function: what did the plan assume, what actually happened, and is the gap timing or direction. Timing gaps close themselves; direction gaps compound. Confusing them is the most common misreading at this stage.
Then run four specific searches, in this order. One: a number that appears differently in two places. Two: a commitment one team is counting on that another has not made. Three: a trend visible only when two functions are read together, such as rising support volume against a shipped feature. Four: an assumption in the plan that is no longer true. These four are searches, not themes, and they are worth running literally every month, because the value comes from the discipline rather than the insight.
Rank findings by consequence, never by novelty. For each, state what it costs if nothing changes, in money or in weeks. An interesting observation with no consequence is a distraction, and including it teaches the chief executive to skim.
Cut to three. One page, three findings, each with the contradiction stated in one sentence, the consequence, and the single person who can resolve it. A fourth finding halves the attention on the first three.
Show every finding to the leader it implicates before it goes to the chief executive. Not for approval, for accuracy, and usually the day before. Two things happen: about a third of findings turn out to have an explanation you did not have, and the ones that survive arrive without an ambush. A synthesis that reaches the chief executive first buys one quarter of drama and costs a year of cooperation.
Land it before the leadership meeting, not in it. Circulated the day before, with the three findings on the agenda as items with owners. Presenting a contradiction cold, in the room, produces defence rather than resolution.
Close the loop. Every finding is tracked to resolved, accepted as a risk, or disproved. A synthesis that raises new things monthly and never closes anything becomes a newsletter, and people stop reading newsletters.
Cull one legacy process a quarter. Ask of every recurring process: if we were starting today at this size, would we do this. Most will not survive the question. Doing this because it got us here is the most common reason growth stalls after this round, and challenging it is specifically this role's job.
Protect a short list of decisions that stay fast. The discipline needed to scale arrives at this stage and quietly kills the pace that got the company here. Write down which decisions are deliberately kept low-approval, and defend them by name when someone proposes a review step.
Synthesis, done properly
The document is not a summary of what each function said. The chief executive already has that, and reproducing it is the single clearest way to make the role look like coordination.
It says: here is what these reports mean together, here is where they contradict each other, and here is the thing nobody is looking at. One page. Three findings. Each with a name against it.
Two habits make it credible. Every number traces to the report it came from, so any leader can check it in a minute. And when a previous month's finding turns out to have been wrong, that is said in the next one, in a line, without elaboration. A document that has never been wrong is not being read carefully by its own author.
Done monthly for a year, this changes how the company is run, and it is the clearest demonstration of the role's value at this stage.
The second management layer
Somewhere in this range the company acquires managers of managers, and the failure is predictable. Directors promoted for being strong managers now manage managers, which is a different job nobody has taught them. Information degrades across two layers, and each layer rounds the news slightly upwards, so what the leadership team believes and what is happening diverge without anyone lying.
Two cheap interventions. Skip levels, structured, on a rhythm, with what is heard written up and fed back to the leader concerned rather than around them. And a written rule about which decisions sit at which level, because the most common complaint at this size is that nobody knows who can decide anything.
If you want to know whether the plan is real, ask three people two layers down what the company is trying to do this quarter. The variance in their answers is a better health measure than any dashboard.
Decide alone, or escalate
Decide alone: the planning process, the cadence, what the board sees and in what shape, how information moves, the sequencing of cross-functional work, and which process gets culled.
Escalate: any trade-off between functions, organisation design, anything about an executive's performance, and anything with a legal or financial commitment.
The judgement specific to this stage is what to put in the synthesis at all. The test: would the chief executive rather hear this from you now or from a customer in six weeks. If the second, it goes in, including when it implicates someone senior and including when you are only reasonably confident. Say the confidence level rather than withholding the finding.
Worked example
Situation. Vantera, workflow software for insurance brokers. Two hundred and ten people, fourteen months after a 46 million Series B. Recurring revenue 21 million against a plan of 34 million by year end. Six functional leaders, a board with two independent directors, and a chief executive who had started saying in one-to-ones that he was being surprised too often.
Task. Produce a monthly view that reconciled the functions, starting within six weeks, without adding reporting work that the leaders would resist.
Action. The first attempt was wrong and is worth recording. It was six pages, organised function by function, and it summarised each leader's report accurately. The chief executive read the first page. The problem was structural rather than editorial: organising by function reproduces exactly the view he already had, and the whole value of the document is in the space between functions.
The second version was one page with three findings, and the first month's set was found by running the four searches against six reports and the plan assumptions.
The first finding was a number appearing differently in two places. Sales reported 4.1 million of new bookings for the quarter; finance reported 3.4. The difference was multi-year contracts counted at total value by sales and at first-year value by finance. Both were defensible, neither was written down, and the board pack had been using whichever number arrived first.
The second was a commitment nobody had made. The sales team had closed eleven deals in the quarter mentioning enterprise permissions in the contract or the security review, with implementation expected by June. The engineering roadmap had that work scheduled to start in October. No individual had promised anything untrue: sales had been told it was on the roadmap, and it was, for October. Estimated exposure was around 2.4 million of contracted revenue at risk of a delivery dispute.
The third was an assumption no longer true. The plan assumed a sales cycle of 61 days, taken from the previous year. The trailing six months ran at 94 days, and the plan's second-half revenue depended on deals that would have had to start before they had.
The finding about permissions was shown to both the sales leader and the engineering leader the day before, which changed it materially: engineering had already resequenced part of the work to July for another customer, so the exposure was smaller than first calculated, closer to 1.5 million, and the fix was narrower.
Result. The permissions work was resequenced to a June partial release covering the eight contracts that specified it, with three customers given a written date in August. The booking definition was settled in one meeting and the board pack was corrected, with a note explaining the change rather than a silent restatement, which the audit-minded independent director later said was the reason he trusted the following pack.
The sales cycle assumption was the largest and slowest to land. It was accepted in month two, disputed in month three when a good month arrived, and settled in month four when the trailing average held at 91 days. The full-year number was reforecast to 27 million in month five, six months before it would have been discovered by missing it, which gave finance time to slow hiring rather than reverse it.
The habit that made the document work was showing every finding to the implicated leader first. In the first six months, four findings out of eighteen were dropped or materially changed after that conversation, which cost some drama and bought a leadership team that stopped treating the synthesis as an audit.
A second scenario, where it goes differently
A company of 160 people, eighteen months after a Series B, where the same method produced a boring document for three months running. All six reports reconciled. The searches returned nothing above a rounding difference, and the temptation was to conclude the company was well run.
The finding was that the reports were being written to reconcile. Two functions had quietly adopted a third's definition of an active account rather than argue about it, and the customer success report was assembled from the same extract finance used, so agreement between them carried no information at all. Independent sources had collapsed into one.
The method changed. Synthesis moved from reading reports to reading the work: sitting in three sprint reviews, listening to twelve support calls, and running skip levels with nine people two layers down. That produced the finding the reports could not, which was that the two largest accounts had both raised the same integration complaint through three different channels and each channel had logged it as a one-off.
What changed the approach was not the company's size. It was that reconciliation only detects contradictions between genuinely independent sources. Where reports agree suspiciously, check whether they share a source before concluding that anything is healthy.
Output
One page, monthly, circulated the day before the leadership meeting.
MONTHLY SYNTHESIS | [month] | Prepared for: [chief executive]
Plan of record: [version and date]
FINDING 1: [one sentence stating the contradiction]
Evidence: [numbers, each with the report and date it came from]
Consequence: [cost in money or weeks if nothing changes]
Confidence: [high / reasonable / early signal]
Resolves by: [one named person] Discussed with them on: [date]
FINDING 2: [as above]
FINDING 3: [as above]
STILL OPEN FROM PREVIOUS MONTHS
| Raised | Finding | Owner | State: resolved / accepted as risk / disproved |
WHAT I WAS WRONG ABOUT LAST MONTH
[One line, or the word: nothing this month.]
Everything longer goes in an appendix nobody is required to read.
Failure modes
Summarising rather than synthesising. Recognise it because the document is organised by function. That structure guarantees it reproduces what the chief executive already has. Organise by finding.
Finding four to eight things. Recognise it by a second page. Rank by consequence and cut to three; the discarded items keep until next month, and most turn out not to matter.
Sending it to the chief executive first. Recognise it when a leader hears about a finding about their function in a meeting. It works once. Afterwards the reports you depend on get carefully written.
Raising and never closing. Recognise it when nobody can say what happened to a finding from two months ago. Fix with the open items table, read out monthly.
Reconciling numbers that are not comparable. Recognise it when a finding evaporates on contact with a definition. Build the metric dictionary first.
Becoming the reporting function. Recognise it when leaders start sending you their numbers instead of publishing them. Synthesis sits on top of functional reporting and must never replace it.
Interesting rather than consequential. Recognise it when a finding has no cost attached. If you cannot state the consequence, it is an observation, and observations go in the appendix.
Edge cases
A leader disputes a finding factually. Check it again the same day. Where they are right, drop it and say so in the next document. Where they are right twice in a quarter, the problem is your inputs, and fixing those beats defending the finding.
The contradiction implicates the chief executive. It still goes in, with the consequence stated plainly and no editorial. Deliver it in person and first. This is the hardest thing the role does at this stage and it is most of what the role is for.
The reports all agree. Check whether they share a source before concluding the company is healthy. Where they do, get an independent reading from the work itself, as in the second scenario.
A reorganisation lands. Synthesis pauses for one month while ownership settles, and the plan assumptions are rewritten before the next one. Reconciling against a plan built for the previous structure produces findings nobody can act on.
The chief executive does not read it. After two months, ask what would make it useful, and be prepared for the answer to be that they want it verbally. Where that is the answer, keep writing the page for yourself and deliver it in ten minutes, because the discipline is in the searches rather than the document.
A finding is time-critical. Anything with a consequence inside two weeks does not wait for the monthly cycle. Send it the day you find it, in three sentences, and note in the next synthesis that it was raised early.
Quality bar
- The synthesis is organised by finding, not by function, and fits on one page.
- Exactly three findings, each with the contradiction in one sentence and a stated consequence in money or weeks.
- Every number traces to the report and date it came from.
- Every finding was shown to the leader it implicates before it reached the chief executive, with the date recorded.
- Open findings from previous months are tracked to resolved, accepted or disproved.
- Where a previous finding was wrong, it is said so in the next document.
- One legacy process has been stopped this quarter, on the evidence of the starting-today question.
- The role owns no function, and everything functional was handed over with a date.
Adapting this to your context
The method was calibrated on venture-funded companies of a hundred to three hundred people with five or six functional leaders. The label is only a proxy.
- "Series B" as a label. It means a leadership team that owns its numbers, two layers between the chief executive and the work, a cadence that survives a busy week, and a board that governs. A professional firm with practice heads, a hospital directorate or a university faculty is here regardless of funding.
- The monthly cycle. Fits a business that closes its books monthly. Where the cycle is a term, a season or a grant year, run the synthesis on that rhythm: a monthly document over a seasonal business finds nothing for months and everything in one.
- Three findings on one page. Sized for one chief executive who reads. Where the principal is a partnership board or a management committee, the three findings need an owner from each part and a longer circulation window.
- The four searches. Written around sales, roadmap and retention. Keep the logic and change the sources: rota against demand, caseload against staffing, enrolment against places.
- What not to change. Every finding is shown to the leader it implicates before it reaches the chief executive, and carries a consequence in money or weeks.
Related skills
cos-at-series-a precedes this and produces the handover that makes synthesis possible; attempting synthesis while still owning functions produces neither. cos-at-series-c follows, when the value moves from reconciliation to carrying a named mandate.
annual-planning-and-headcount produces the plan of record that findings are read against. board-and-investor-management and board-deck consume the synthesis for the board pack. revenue-forecast and weekly-status-update supply two of the inputs. decision-memo is the format for any finding that requires a decision rather than an adjustment. structured-problem-solving is the method when a finding turns out to need real analysis rather than reconciliation.