The limbs have to move in time.
Six years building an industrial software company taught me things the decade of banking before it could not, and nearly all of them reduce to one thing. A company is a set of moving parts that have to be in time with each other — strategy, product, sales motion, hiring, budget, funding. Any two out of phase and the rest cannot compensate. It will crawl before it walks; that is normal. But the limbs still have to move together.
None of what follows was learned somewhere large enough to absorb its own mistakes. That is rather the point — where nothing is absorbed, you see the mechanism.
The client decides it. Nobody else gets a vote.
The company had built something genuinely impressive: a robot cell that could be programmed for a new odd-form pick-and-place task in well under a minute, where the industry norm was measured in days. Investors loved the demonstration — start your clock, we will have it running in thirty seconds.
The unit economics did not work, and the reason was not obvious from the inside. The addressable job sat in the seam between volume and variation: enough volume to justify a station costing six figures before any software margin, and a mix stable enough that the programming advantage stopped being the point. Those are precisely the jobs that thirty years of outsourcing had already moved to low-cost countries. Manufacturers also do not evaluate in percent, they evaluate in years — nobody buys a payback they cannot see inside two or three, and against an operator wage of a few dollars a week the arithmetic never closed. Underneath all of it sat a supply chain of robots, tables, grippers, feeders and spare parts — the kind of thing an equity story rarely dwells on and a balance sheet cannot ignore.
So the robots were killed. It was the right decision and not a comfortable one: it meant telling investors that the thing they had funded and admired was not the business. What the episode settled, permanently, is the sentence I have repeated most often since — product-market fit is not determined on a spreadsheet, in a slide deck, or by investor enthusiasm. It is determined when the product meets the client, and the numbers have to work as well as the technology does.
You cannot start at the top of the market. You climb to it.
What replaced the robots was a system nobody in the company had a name for yet — it turned out to be a manufacturing-execution system, a category none of us had come from. After the first deployment, sold unfinished to a battery manufacturer against vendors from the analyst quadrants, the useful question was not how we won. It was which parts of that win generalise. The first version was one to be embarrassed about, which is the correct condition for a first version.
The answer became a strategy document in 2022 that held: manual-assembly-heavy manufacturers building complex products through complex processes, carrying real quality, documentation and traceability obligations. Automotive, aerospace and defence, power infrastructure, building systems.
The acquisition sequence mattered as much as the target. A defence prime will not put an untested vendor inside its production system, so starting there wastes years. What worked went in three steps: first, fast-moving startups with high risk tolerance and no legacy to protect; then established mid-market manufacturers with brands they could not afford to put at risk; and only then the primes and the large infrastructure groups. Each tier is the reference that makes the next one possible. Credibility of that kind cannot be bought and the steps cannot be skipped.
Some systems are open-heart surgery for the business that buys them.
The system that runs a factory floor. The platform that moves a large corporate's cash. The core record a business cannot operate without for a single day. These are not products, they are one-way doors — the cost of getting the choice wrong exceeds the licence fee by orders of magnitude, which means the decision is about almost everything except money.
That inverts the sales motion. It is a risk decision rather than an opportunity decision, so the buyer is not asking what they might gain, they are asking what could go wrong and who carries it. Many stakeholders, each with a different fear, over a cycle measured in years. It is never about price. Being the smallest vendor in the room makes all of it harder, and it changes tactics more than strategy — you cannot win on institutional comfort, so you win on being unmistakably right about the customer's own process.
The useful part was recognising the shape rather than learning it. I had spent a decade on the bank's side of exactly this purchase, watching corporates choose infrastructure they could not easily unwind. Different industry, identical anatomy. That transfer was worth more than sector expertise would have been, and it is why I recognise the shape of that purchase quickly now.
Find out why they actually bought. It is rarely what you assumed.
I have won deals I was not the favourite to win, in banking and in software both. Where the relationship later allowed the question, I asked the client directly: why us? The answers were seldom the ones in our own sales material.
The specification has to be right — that is the entry ticket, and there is no version of this where being wrong about the customer's process survives contact. But once several vendors have cleared technical diligence, the remaining question is not a technical one. It is who the buyer will be speaking to every week for the next decade. Signing that order is a commitment to a relationship rather than to a feature list, and the buyer knows it even when nobody says so out loud.
That is the commercial case for culture, and it is not a soft one. Culture is what the customer is actually underwriting once the specification has stopped being a differentiator. For shorter-lived and more transactional products the same function tends to be served by brand — a promise, made in advance, about what dealing with you will be like.
The discipline is in the asking. Go into those conversations without a preferred answer: a buyer will happily confirm whatever you appear to want to hear, and a flattering explanation for a win costs more than an uncomfortable one, because it teaches you to repeat the wrong thing. Both halves — the technical and the cultural — can be built deliberately, but only once you know which of them is carrying you.
Strategy, sales cycle, budget, funding cycle — one clock.
This is the lesson I would write on a wall. The strategy has to match the sales cycle, which has to match the budget, which has to match the funding cycle. Break the chain at any link and the other three cannot compensate for it.
The failure is arithmetic rather than character. When the sales cycle is materially longer than the funding cycle, a company is structurally short of time — and no amount of effort inside the period closes a gap that was set when the plan was written.
The corollary is about capital, and it is the more useful half. Investors are not interchangeable. One who understands a long, risk-based enterprise cycle and can live with its rhythm is worth more than one offering a better price, because they also know what sits on the other side of it: a product that is extremely hard to displace, high gross margins, and a great deal of area under the curve. Matching the investor to the deal motion is part of designing the company, not a separate exercise in fundraising.
Manufacturing is much harder than most give it credit for.
I had financed manufacturers for a decade without once thinking of manufacturing as a hard problem. That was my error, and I do not think it was an unusual one: from the capital side a factory looks like a solved problem with a cost curve attached. It is not. How product flows through a plant, how internal logistics actually work, what a given layout does to working capital, throughput, quality and the health of the people doing the work — that is deep, hard-won expertise, and it tends to sit with people who are not in the room when the capital is allocated. Manufacturing is badly under-rated as a source of both value creation and value destruction, and the gap is usually one of distance rather than ability.
The version I use to explain it: next time you are in a McDonald's, consider that the menu permits an enormous number of distinct orders, and any one of them arrives hot in about three minutes. That is a car plant. Different scale, different stakes, the same flow mechanics and the same synchronisation problem. Most of it transfers directly to how work moves through a software or services business, which is the second reason it was worth learning properly.
The digitisation layer is only worth something on top of that underlying logic. Knowing the true cost of building and implementing these systems — and the real buy-versus-build arithmetic, which agentic AI is genuinely moving — is what brings it into 2026.
Direction is a kindness
A clearly articulated goal lets people make good micro-decisions without asking. Without it they hesitate, and hesitation reads as incapacity when it is actually an absence of instruction. The military understood this better than I did at the time.
And the truth has to come back up
Intent travelling down is only half of it. What comes back up is filtered toward what people believe you want to hear — not dishonestly, just rounded. The distance between you and the floor is set by what an organisation treats as authoritative, not by its size or anyone's title. At essay length →
Measure so you can reverse early
KPIs, OKRs, whatever they are called — their real function is to make a bad decision, including a bad hire, reversible in weeks rather than defended for a year. They only work if they are tracked and genuinely tied to the goal.
Long cycles need a different seller
Working warm leads to a script and feeling your way through an eighteen-month, risk-based, multi-stakeholder cycle are different crafts. The second matters more than subject-matter expertise, and experience of a full formal procurement is rarer than it looks on a CV — which makes it hard to screen for, and worth screening for.
Coach, and be present
I under-invested in this. The transition from selling yourself to having others sell is expensive in time and money when the subject matter is complex, and a manager who is on the road is not coaching. It needs a plan and a budget rather than a hope, and I would give it both next time.
Rooms with big titles
They are just people. Understand how they are measured and work with it. Never make them look bad. And do not engage unless the stakes are large enough to matter to them — what is existential to a small company is often a rounding error to a large one.
You cannot build a future on a past success.
This holds equally for a first small win and for a century of them. A victory that makes a team complacent and an institution's long record of having been right produce the same effect: the assumptions stop being examined. The day that happens is the first day of the decline, and nothing announces it.
It is also why synchronisation is not a state you arrive at. The arrangement decays the moment nobody is re-tuning it, and past success is the most reliable reason to stop re-tuning. Markets move, buyers change, costs shift — an arrangement that was in phase last year is not automatically in phase this one.
What keeps it alive is appetite, and appetite is specific: wanting to be a little better than you were yesterday, hating to lose, genuinely caring about basis points of margin — most of all the ones that came from deliberate effort rather than luck. Without that, an organisation is not really a company at any scale. It is a project. The two look similar for a while and end very differently.
The leader's job is not to carry that appetite alone but to make other people want it: celebrating wins properly, dissecting losses without flinching and without blaming, and then adjusting — a shorter loop, a cleaner process, a better answer than last quarter's. Everything else on this page is technique. This is the part that has to be true first.
None of this is a theory of management. It is a diagnostic habit: when something in a company is not working, the first question I now ask is which two parts have fallen out of time with each other — because it is rarely one broken limb and almost always a timing problem between two working ones. A strategy the budget does not fund. A hiring plan the sales cycle cannot carry. A product the market evaluates on a horizon nobody modelled. Getting those into time with each other is most of the job of running a company, and most of what an investor is underwriting when they back one.