People & Structure
Who owns what, which decisions can be made without you, and what happens when someone is out for a week.
The Guide
What breaks first.
An operational timeline for healthtech and medtech companies, ten people to sixty. What fails at each stage, in what order, and why none of it is bad luck.
5headcount bands
4categories it lands in
12 minto read
Operational failures arrive in an order, and the order is set by headcount.
Founders tend to experience operational failure as a surprise. A customer escalates. A hire doesn’t work out. A diligence request lands and takes three weeks to answer. Each one arrives on its own, feels like bad luck, and gets handled.
Headcount sets their order more than revenue, funding stage, or product quality does. A company of eighteen people breaks in a particular way. So does a company of forty. Each failure means the company outgrew a structure that used to fit, and nobody did anything wrong.
That’s the useful part. If the sequence is predictable, you can see the next one coming.
What follows is that sequence, roughly by headcount, for healthtech and medtech companies from Seed through Series B. The bands are approximate. Companies don’t break on a specific hire, and yours might run five people ahead or behind. The order holds better than the numbers do.
Every stage below is the natural consequence of a structure that worked at the previous size. The companies that struggle most are usually the ones that were running well enough that nobody looked up.
The founder is the operating system, and that’s correct.
At this size, running everything through one person is the right answer. Coordination costs are real, and a company of eight needs a founder who knows everything and decides quickly. Adding structure here is a waste.
Most operational advice is written as though process is always good. At eight people, the founder holding it all is the most efficient arrangement available, and a founder who feels guilty about that is feeling guilty about the wrong thing.
At this stage, the arrangement only has to be deliberate. There’s a difference between a founder who knows the company runs on them and a founder who assumes it runs on systems that don’t exist.
The sign this stage is ending: you start hearing about decisions after they’ve been made, and they’re mostly fine. It means the queue behind you has gotten long enough for people to route around it.
The founder’s calendar becomes the constraint.
This is where the arrangement that was efficient stops being efficient, and it happens quietly.
The mechanism is simple. Decisions per week grow with headcount, but the founder’s hours don’t. At some point the number of decisions that need the founder exceeds the number the founder can make well. Then people stop asking.
They stop because asking means waiting, and waiting means missing their own deadline. So they guess. The guesses are mostly right, which is exactly why this doesn’t surface as a problem. Nobody escalates a decision that turned out fine.
The cost accrues in two places. First, a small percentage of those guesses are wrong in ways that show up months later, usually in a customer relationship or a commitment nobody wrote down. Second, and more expensive, your best people learn that you don’t trust them to decide. That’s the lesson they take, regardless of what you intended, and it’s the one that eventually shows up as attrition.
What’s usually missing: any written statement of what people can decide without you. An approval matrix is too heavy for twenty people; a clear answer to “what am I allowed to call?” is enough.
The healthtech wrinkle: in a regulated business, some decisions do need to route through one accountable person, and founders at this stage often can’t articulate which ones. So they either hold everything, which creates the queue, or they delegate uniformly, which puts regulatory judgment in the hands of someone who wasn’t told it was regulatory. Both are avoidable, and neither is avoided by accident.
The sign: ask yourself how long it’s been since you took a full week off. If the honest answer is that it feels impossible rather than merely inconvenient, you are in this band, and probably past the middle of it.
The management layer nobody built.
Somewhere in here, the org chart quietly becomes a hub and spokes, with eight or ten people on the spokes, all reporting to the founder, all needing something.
This is the most common structural problem I see, and it’s the one founders resist longest, for a reason that’s worth respecting: promoting someone into a management layer means putting distance between you and the work, and at this size the work is still the thing you’re best at.
But the arithmetic doesn’t care. Ten direct reports means either ten shallow relationships or a founder doing nothing but one-on-ones. Most founders pick the first without deciding to.
Two other things surface in this band, both of them consequences of the same gap.
Hiring stops being repeatable. Every search starts from scratch, because there’s no scorecard, no defined interview loop, and no agreed standard. The first bad hire usually lands here, and it’s expensive twice: once in salary and severance, and once in the six months the team spent working around the person before anyone said it out loud.
Onboarding depends on a person. A new hire’s ramp depends entirely on how much time someone senior can give them, which means ramp time is inversely proportional to how busy the company is. New people arrive fastest when they’re needed least.
What’s usually missing: a real leadership layer with decision rights, and a written hiring process that survives the founder not being in the room.
The sign: count your direct reports, then count how many of them need a decision from you in a normal week. Six or seven self-sufficient senior people is a workable span. Six who each need something from you is a queue, and the queue is the management layer you haven’t built.
Compliance and finance stop being survivable on goodwill.
This is the band where the outside world starts setting your operational deadlines, and it’s the one that costs the most when it’s missed.
Two things converge. Enterprise customers arrive, and before they buy, they send a security questionnaire, ask about data handling, and want to know who owns HIPAA and what happens when someone leaves. Separately, if you’re raising a Series B, the diligence is a different animal from the seed round. Nobody is buying the vision anymore. They’re testing whether the numbers hold.
The timing is what makes this expensive.
Compliance built under a customer’s deadline costs several times what the same compliance costs built on purpose, and most of the difference is attention. A deadline-driven compliance build pulls your senior technical people off the roadmap for a quarter, during the exact quarter you were counting on them to ship. You pay for it twice, and the second payment is the one that shows up in your board deck.
The financial half has the same shape. Reporting that was adequate for a seed deck fails a real diligence process, usually because the numbers can’t be produced quickly or traced to anything. Founders in this position usually know their business cold and can’t demonstrate it fast, and hesitation in a term sheet negotiation gets read as a signal about the company rather than about the reporting.
What’s usually missing: a named owner for compliance whose job description includes it, and financial reporting that a stranger could follow without a narrator.
The healthtech and medtech wrinkle, and it’s the important one: your regulated surface arrives earlier and from further outside than it does in other categories. A B2B software company gets asked about SOC 2 when it moves upmarket. You get asked about HIPAA, data governance, and often FDA-adjacent questions by your second real customer, because your buyer is a health system or a payer with its own regulator sitting behind it. That compresses the timeline. The band where most companies can still improvise is a band where you usually can’t.
The sign: imagine an enterprise customer sends a security questionnaire on a Friday afternoon. If the honest answer is that it becomes your weekend, you’re in this band.
The seams.
By this size the company is too large to overhear itself, and that changes what fails.
Up to about forty people, a surprising amount of coordination happens by proximity. People catch things in passing, in a shared channel, or because they happened to be in the room. It doesn’t look like a system, so nobody notices it’s load-bearing. Then the company crosses into a majority of people who weren’t there for the founding conversations, and the informal layer stops carrying the weight.
The failures move to the seams. A handoff between two teams. A commitment made to a customer that nobody wrote down. A launch that slipped because the dependency was obvious to everyone except the person who owned it. Individually each one is small and gets fixed. Collectively they are the reason the company feels slower at fifty people than it did at twenty-five, with no single decision you could point to.
Tribal knowledge turns from a quirk into a liability in this band. At fifteen people, one person knowing how something works is fine. At fifty, it’s a single point of failure with a vacation schedule.
What’s usually missing: written process at exactly the seams that break, which is a much smaller set than “document everything,” and a reporting rhythm that surfaces a dropped ball before a customer does.
The healthtech wrinkle: you now have clinical or scientific staff and commercial staff in the same company, and they have different operating cultures. One is trained to be careful and slow on purpose, because being fast in a clinical context is how people get hurt. The other is measured on speed. Both are right for their work, and a cadence designed for one gets quietly ignored by the other.
The sign: ask how you usually find out something fell through the cracks. If the answer is that a customer or an investor tells you, the seams are already carrying more than they can hold.
These stages depend on each other, and only in one direction.
It’s tempting to read the above and start on whichever one sounds most alarming. That’s usually a mistake.
Compliance built before there’s a management layer has no owner, so it becomes a binder nobody maintains, which is worse than not having it, because now you believe you’re covered. Financial reporting built before there’s an operating cadence gets produced on time and read by nobody. Process written before the seams exist documents a company you no longer have by the time the ink dries.
Decision rights make a management layer possible. A management layer gives compliance and finance somewhere to live. Both of those give you something real to report on. And a reporting rhythm is what makes seam failures visible before customers find them.
Skipping a stage produces a specific frustration: you built something and got no benefit from it.
Everything above lands in one of four categories.
The frame is useful mostly because it stops the conversation from becoming “everything is broken,” which is never true and never actionable.
Process & Cadence
The meetings and reporting that keep the company on plan, what’s written down, and where handoffs drop.
Compliance & Risk
What obligations apply today, which ones are coming, who owns them, and what happens when someone outside asks for proof.
Financial Infrastructure
What reporting exists, whether it survives scrutiny, and how many people can explain the burn rate.
All four are measured the same way: by what happens to them when you’re not in the room.
People and structure is decisions only you can make. Process and cadence is knowledge only you carry. Compliance and risk is obligations only you are tracking. Financial infrastructure is numbers only you understand.
At this stage, founder dependency isn’t one problem among four. It’s the same problem, showing up in four places.
Four questions that place you better than a headcount does.
They’ll tell you which band you’re in, and it’s usually a different one than founders guess.
How long has it been since you took a full week off?
A week without your laptop. If it feels impossible rather than inconvenient, you’re at least in the 10 to 20 band, whatever your headcount says.
How many of your direct reports need a decision from you in a normal week?
The raw count matters less than that number. Seven senior people who run their own areas is fine. Four who each need you to unblock something is a queue with your name on it.
If an enterprise customer sent a security questionnaire this afternoon, what would happen?
If the answer involves your weekend, the outside world is already setting your operational deadlines.
How do you usually find out something fell through the cracks?
If a customer or an investor tells you, the informal layer has stopped working and nothing has replaced it.
Less than you might think, immediately.
The most common error I see is a founder reading something like this, recognizing three of the bands at once, and starting a broad operational overhaul in the same quarter they were supposed to be shipping. That reliably produces a lot of half-built infrastructure and a delayed roadmap.
The more useful next step is smaller: find out which one is furthest behind, and which one reaches you first. Those are two different questions, and the second one matters more. A significant gap in a category nobody outside your company will inspect this year is less urgent than a moderate gap in one your next customer will inspect next month.
That’s a diagnosis, and it’s difficult to do on yourself, for the same reason it’s hard to proofread your own writing. You know what you meant.
I built a short version of the diagnosis you can run yourself. Twelve questions across the four categories above, about five minutes, and it gives you four scores and an honest read on which one is furthest behind. It won’t tell you what to do about it, because twelve questions can’t, and anyone who claims otherwise is selling you a template.
If you’d rather talk it through, that’s thirty minutes, and there’s no charge. Bring the current state of your company, and we’ll work out together whether anything here is worth doing. At the end I’ll tell you what I think, including the possibility that the answer is to do nothing right now.
Which one is furthest behind at your company?
Twelve questions across the four categories above, and about five minutes. Four scores, and an honest read on which one reaches you first.