The message lands at 9:40 on a Tuesday night. Why is acquisition cost up this week.
Nothing is actually wrong. Cost per customer moves week to week the way weather does, and the honest answer is that seven days of it is noise.
The question still costs a day and a half. Pull the numbers. Write the explanation. Walk the founder through it on a call neither calendar had room for, then rebuild the week the interruption ate.
I have been the first marketing executive a company ever hired, more than once. The founder in that seat is running a weekly experiment on you and never mentions the design.
A founder new to having a CMO is really deciding one thing: whether they can stop checking your work.
Most operators answer that experiment with output. More work, better work, work delivered faster. It rarely registers, because the founder has no format in which to read any of it.
Why competence alone buys you no room
A founder who has hired engineers knows what supervision looks like there. Code ships or it does not, and the calendar tells you which.
Sales is easier still. Pipeline fills or empties, deals close or stall, and the scoreboard updates itself in public without anyone having to ask.
Marketing offers neither tell. Money leaves the account continuously and results arrive on a lag, so a founder watching in real time sees only the outflow.
So they reach for the one instrument they have, which is asking. Each question is cheap to send and expensive to answer, and it returns a snapshot with no baseline attached, which produces another question seven days later.
That is how a relationship defaults to surveillance without anybody choosing it. Nobody announces that trust is low. The engagement simply accumulates interruptions until good work starts reading as opaque spend.
Engagements rarely die because the marketing was wrong. They die because a founder ran out of patience for not knowing, usually somewhere in month four. I have written elsewhere about what a fractional engagement is supposed to deliver, and this is the part that quietly decides whether it lasts long enough to deliver any of it.
The contract goes in during week one
The first artifact I build in an engagement is a reporting contract. It gets agreed in the opening week, while there is nothing good to report and nothing yet to defend.
It fixes four things: the day, the shape, the numbers on top, and who hears what first. Almost everything else about an engagement changes. These do not.
One update, Monday morning, before the founder's week has taken a shape. The same four blocks in the same order, whether the week was excellent or ugly.
The constancy is the mechanism. A document that changes shape has to be read from the top every time, and a founder re-reading a report from the top is still doing the checking themselves.
A fixed format gets skimmed in under a minute. After a month the founder stops parsing structure and starts noticing deltas, which is the moment the update begins doing actual work on their behalf.
Three numbers, and the same three every week
The scoreboard is where operators overreach. You have forty numbers, the work behind them was real, and the instinct is to prove it with all of them at once.
A founder can hold three. Choose the three that would change a decision if they moved: one demand number, one efficiency number, one money number. Which three depends on the business, and I go through how I pick them in the three metrics I want every founder watching.
Then leave them alone. Swapping the scoreboard resets a founder's memory to zero and they will not tell you that it did. When the three genuinely have to change, say so in the update and run the old set alongside the new one for a month.
None of it survives numbers that cannot be defended. At FX Replay the reporting only became possible once the measurement layer had been rebuilt: a single dictionary of four event definitions, fired from the backend so every system recorded the same moment. Before that, three tools gave three answers.
No scoreboard outlives a founder discovering that. Honest measurement has to come first for exactly this reason, because every number stacked above it inherits whatever the measurement gets wrong. The growth operating system is where I lay that sequencing out in full.
The misses go first
The default report is chronological, or flattering. Wins on top, the problem buried in the sixth paragraph, where an anxious reader finds it anyway at the worst possible moment.
Invert it.
Send the bad number before they find it. A report that names its own miss buys more autonomy than a report that hides it.
The diagnosis is the load-bearing half. A miss with no explanation is an alarm, and a founder hearing an alarm has no option except to investigate it themselves.
A miss with a cause and a dated fix is a problem you have already taken off their desk. Same number, opposite effect on the person reading it.
So the rule gets said out loud in the first week, in plain words: you will never learn a bad number from a dashboard before you learn it from me. A founder who finds a problem first has to pick between two explanations, hiding or not noticing, and both are worse than the number.
Leverage Companies had never employed a marketing executive before me, and one of the earliest documents I wrote for the function there was a set of team expectations. It said that when a deadline is going to slip, say so early. Holding a team to a standard I broke upward would have cost me both the team and the room.
Decisions travel with a recommendation attached
The other half of the update is the ask, and this is where polite operators do real damage. Should we shift spend out of display is a reasonable question that a busy founder has no way to act on.
Decisions travel up with a recommendation attached, or they come back down as homework.
The format is four parts in one line: the call, what I advise, the one-sentence reason, the date. Move $12k out of display into search, because display has not cleared payback in six weeks, and if I hear nothing by Friday it moves Monday.
That last clause carries most of the weight. It turns silence into a decision, so a founder who is busy or travelling never becomes the thing the work is waiting on.
It also puts the risk where it belongs. You own whatever happens when nobody replies, which is a strong incentive to write the clause only when you are genuinely confident.
The deadline has to be real, too. Miss your own Friday twice and the founder learns that the whole update is theater, and every check-in you had retired comes straight back.
The engine underneath and the surface on top
None of this is the meeting where the work actually gets decided. The weekly growth review is internal: the team, one dashboard, experiment readouts, and the hour where money moves between channels.
The founder update is the outward face of that engine. One page, three numbers, the misses, and only the decisions that need someone above me. Forward the review agenda instead and you have handed a founder forty numbers with no argument attached.
The two connect at the misses. Every miss in the founder update already has a name against it inside the review, because the function is built so each seat carries a metric of its own. That sequencing is what I use when building a growth team from nothing, and it is what keeps the founder update honest rather than decorative.
What thirty days of this buys
Weeks one and two feel like nothing at all. The founder keeps checking, the update goes out anyway, and you are paying into a reputation the format has had no time to earn.
The inflection arrives at the first bad week. The founder opens the update and finds the problem sitting at the top of it, diagnosed, with a fix already dated.
They watched the report do, unprompted, the job they had been doing by hand on Tuesday nights. That is the week the checking becomes optional, and it has almost nothing to do with the marketing.
From there the check-ins thin out quickly. By the end of the month the questions have changed character: they stop being requests for status and start being opinions about the recommendation, which is a founder deciding rather than supervising.
That is the trade the contract makes. You give up control over what gets seen, including the parts you would rather explain in person first, and what comes back is room to operate. Autonomy gets handed to the operator who made themselves auditable, and to nobody else.
Install it before you need it
If you are four months into an engagement and the founder is still checking, more results will not fix it. Results have been arriving the whole time. What is missing is a shape the founder can read them in.
The underlying idea is old. Gabarro and Kotter, writing in 1980, defined managing your boss as "consciously working with your superior to obtain the best possible results", which reads as obvious until you notice how few operators treat it as work with a format.
What is different in a fractional seat is the clock. You have weeks where a full-time executive would have years, and the person above you has never supervised this function in their life.
The contract is cheap to install and expensive to retrofit. Write it on day three, before you have results worth protecting, and the first hard month arrives with the format already load-bearing.
If you have brought in your first marketing executive and the reporting still feels like an interrogation, this is the kind of contract I install. Let's talk.