Functional vs. Divisional Structure for Growing Companies
Somewhere past a hundred people, someone will propose breaking the company into units — by product, by segment, by geography — each with its own engineering, product, and go-to-market. It is usually proposed as a solution to slowness, and it is sometimes correct. It is also one of the most expensive structural mistakes available when done early, because it is very difficult to reverse and it duplicates cost permanently. This page is about the preconditions, and about the middle path most companies should take instead.
What’s inside
The two structures answer different questions
What divisional actually means
Four preconditions
What divisionalising costs you
The premature-division failure
The middle path
Where Blomma fits
The two structures answer different questions
Functional and divisional are not two points on a spectrum of maturity. They optimise for different things.
Functional structure groups by craft— all engineers together, all marketers together. It optimises for depth, consistency, and efficient use of specialists. Its cost is that anything requiring several functions requires several leaders to agree, so cross-functional work is slow.
Divisional structure groups by outcome— each unit contains the functions it needs to serve its market. It optimises for autonomy, speed within a unit, and clean accountability with its own numbers. Its cost is duplication, drifting standards, and shared platform work that no unit will fund.
So the question is not which is more grown-up. It is whether your bottleneck is cross-functional coordination or craft consistency. If you are slow because every decision needs three functions, divisional relieves that. If your quality varies by team and nobody can say what good looks like, divisional makes it considerably worse.
Worth noting: most companies under two hundred people who think they need divisions actually need clearer decision rights inside a functional structure. That is a much cheaper fix and it should be tried first.
What divisional actually means
The word gets used loosely, which is where the trouble starts. A genuine division has four properties.
Its own P&L, or something close to it.Revenue and cost attributable to the unit, so accountability is real rather than notional.
Enough function inside it to operate.Its own product and engineering at minimum, usually its own go-to-market. A “division” that has to request engineering from a central pool is not a division; it is a product team with a grander name.
Real decision authority.The unit leader can set roadmap, hire within budget, and make pricing calls in their market without escalating. If they cannot, you have added a layer and no speed.
A distinct market or customer.Two units serving the same customer with different products will conflict continuously over the relationship, and the conflict will route to you.
If a proposed division lacks any of these four, it is not going to deliver the autonomy that justifies the cost. Test the proposal against all four before agreeing to it.
Four preconditions
Divisionalise only when all four hold. Founders typically have two and proceed.
One. Genuinely distinct markets or products.Distinct enough that the right roadmap, pricing, and go-to-market motion differ materially. Two products sold to the same buyer through the same motion do not need separate units — they need a clear owner each inside one structure.
Two. Enough scale to duplicate without crippling either unit.Each division needs a credible functional bench. Splitting twelve engineers into two groups of six usually produces two teams that can no longer do platform work rather than two autonomous units.
Three. A unit leader who can genuinely run it.This is the binding constraint in practice. A division requires someone who can own an outcome across functions — effectively a general manager. If you do not have that person, you are creating a role you will fill badly, and the failure will be attributed to the structure.
Four. A settled answer on shared platform and services.What stays central, who funds it, and how units get served. Unresolved, this becomes the dominant source of conflict within a quarter, and it escalates to you.
What divisionalising costs you
Worth pricing explicitly, because the proposal usually arrives with the benefits itemised and the costs implied.
Duplication, permanently.Two of several roles you previously had one of. Sometimes worth it; always real.
Standards drift.Each unit develops its own conventions, tooling, and quality bar. Within a year the product feels assembled from parts, and metric definitions diverge until two units report different numbers for the same thing.
Platform underinvestment.Shared infrastructure benefits everyone and is nobody’s priority. Unless you fund it centrally and explicitly, it decays.
Career-path narrowing.Specialists in a division have fewer senior peers and a shorter ladder, which is how you lose your strongest specialists eighteen months after a reorg nobody connects to their departure.
Culture divergence.You stop having one company and start having several, one per unit, unless someone actively holds the line — and that someone can only be you.
The premature-division failure
The most common version of getting this wrong, worth recognising because it has a specific shape.
A company is slow because decisions require several functions to agree and nobody has clear authority. Someone proposes divisions as the fix. The company divides, and for two quarters things feel faster — because the reorg reset priorities and everyone is newly focused, not because the structure is better.
Then the costs arrive. Platform work has stopped. Two units are competing for the same customer. Standards have diverged. The units are too small to be autonomous, so they are requesting resource from each other and escalating when refused. And the original problem — unclear decision rights — has reappeared inside each unit, because it was never addressed.
At which point the options are recentralising, which is a second reorg and a serious hit to credibility, or living with it. Most companies live with it.
The test that prevents this: write down the three decisions that are currently slowest, and check whether divisionalising actually relocates them. If those decisions would still require the same people to agree, the structure is not the problem.
The middle path
For most companies between roughly eighty and two hundred and fifty people, the right answer is neither — it is a functional structure with genuine outcome ownership inside it.
Concretely: keep functions for craft, standards, and career development. Assign named owners to outcomes that cut across them, with real decision authority over their outcome — roadmap, priority, and the trade-offs inside their scope. Give each outcome owner a small dedicated group where the work is tightly coupled, and draw on functional depth for the rest.
This gets you most of the speed benefit of divisions without the duplication or the standards drift, and it is reversible. It also surfaces whether you have anyone capable of running a division, because outcome ownership is the training ground for exactly that role. A company that cannot find good outcome owners inside a functional structure is not ready to divisionalise.
Then divisionalise later, when the four preconditions genuinely hold, with a leader who has already proven they can own an outcome.
Where Blomma fits
This decision is usually argued by someone who wants to run a division, which is not a criticism — ambition is what you want in senior people — but it does mean the proposal arrives with a position attached. Your other executives will have views shaped by whether the split advantages them. You are the only person in the room without a stake, except for whatever you absorbed from wherever you saw this done before.
Blomma is an always-on AI career coach with no stake in your structure. Use it to test the four preconditions honestly, particularly the third — whether you actually have a general manager or are about to promote someone into a role they are not ready for. Use it to run the slowest-decisions test and check whether division relocates them. Use it to price the costs explicitly before you agree. And use it to prepare the conversation with the executive whose proposal you decline, or who does not get the unit they wanted, which is the part of this decision that determines whether they stay..
Divisionalising is a real and sometimes necessary step. It is also close to irreversible, expensive in ways that arrive late, and frequently proposed as a fix for a problem that clearer decision rights would solve for nothing.