The arithmetic of leverage, applied to how your business runs.
We redesign how work gets done — sharper processes, clearer decisions, and systems that scale. We bring modern tools, AI among them, to bear exactly where they earn their place, and nowhere they don't.
Start with the symptomBefore the solution
Nor is “our forecast is bad” usually a forecasting problem. You already know the symptom. It is rarely the problem.
By the time anyone names it, the business has organized itself around it. A spreadsheet exists to work around it. A step in the process exists only because of it. Nobody remembers why any more — it is simply how the work gets done.
So we open with a differential, not a proposal: the competing causes, and the one cheap test that tells them apart. Below are the twelve most common, and the test we would run for each. All twelve, in full. No form.
Pick the one you have said out loud in the last month
Why waiting is expensive
A problem caught early is a policy change — an owner and a Tuesday. Left alone it becomes a systems project, then a reorganization, then a capital case. Roughly a class per year. Nothing about the defect moved; the workarounds around it did.
The stages are fixed; only the methods inside them change. That is what lets one small firm work across industries without pretending to industry tenure — the complaint differs, the procedure does not.
Never skip to call three. An optimizer aimed at a misdiagnosed problem returns a precise wrong answer, delivered with confidence. It is the most expensive mistake available and the easiest one to make.
One lever moves one quantity. The quantity moves the identity. The identity is enterprise value. Nothing in between is a matter of opinion — which is why the same engine can start at a changeover time and end at a price per share.
3.3% return on the capital employed, against a 9% cost of capital — −$44.7m of value leaving every year, reported nowhere. That is the shape we find most often.
ARCOVA is calls two, three and four running as software. It reads the filings, finds where value is leaking, prices every action against the cost of capital, sequences them, and then learns from what actually happened — which feeds straight back into the model and changes what it recommends next. You can argue with it in the room and it will recompute while you do. The screens below are the real product, on an illustrative company.
Every finding names the disclosure it came from and states what it permits you to conclude — and, just as usefully, what it does not. Modeled figures elsewhere are constrained by these and cite them, so nothing is asserted that cannot be walked back to a line in a document.
Where two sources disagree, the conflict is declared and the position we took is recorded. A skeptic should find our weakest link faster than we could hide it.
The benchmark is the cost of capital, not last year. A business can be comfortably profitable and still destroying value, and only one of those two facts shows up in the management accounts. Every leak is sized, sourced and tied to the node it comes from.
Identified leakage deliberately exceeds the gap that has to be closed. Not all of it is recoverable, and the overlaps are declared rather than quietly summed.
Steps three to six are a cycle, not a sequence. Each result you report re-estimates the tree, which changes the best combination, which reshapes the schedule, which sets up the next thing to measure. The model of your business is worth more in year two than it was on the day we handed it over — which is the opposite of how a deck ages.
Any model can produce a number. The harder discipline is declining to, and each of these is a rule in the code with a source behind it.
There is no public login and no self-serve trial. We build your profile from your filings before the first meeting, then run the engine live — your business, your constraints, every figure traceable to the document it came from. The screens above are a demonstration company; nothing on this page is real.
Ask for the first readEnumeration is arithmetic, and arithmetic is reproducible. What is not reproducible is the content it operates on. Before any number is computed, every action carries four things written by hand: the thesis that makes it plausible, the test that would settle it, the way it fails, and the source the rule came from. That last set is the part that took years, and the part a competitor cannot generate.
And it does not rank the actions and take the top few — it scores every combination of them. Ranking assumes the actions add up. They do not: relieve one constraint and the next one binds, and two fixes drawing on the same scarce attention deliver less than either promised alone.
What it reads, the eight stages it runs, the eight levers it moves, the nine analyses it returns, and the things it refuses to compute. Everything the engine does, stated so it can be argued with rather than admired.
Open the fact sheet Access phrase required — ask and we will send it.
We hunt for the small, well-placed changes that produce disproportionate results — and we prove them before we build them. A process is a network. The work is finding the one point where a small, precise change moves the whole outcome.
Convexity tells you what to change. Concavity tells you what order to change it in — and the escalation curve above is what the wrong order costs while you wait.
Nobody should have to commit to four weeks of consulting to find out whether a consultant is right. So we do the first piece of work before you engage us, on public filings, at our own cost — and you see a real number about your own business before anything is agreed.
We have also run this on named public companies, from their filings alone. Those case studies are not published — a live company deserves to hear a conclusion about itself before the internet does. We will walk you through one.
Ask for the first readThe twelve above turn up in plants, in distribution networks, in capital programs and in professional firms — described in completely different words each time. We work wherever a decision is data-rich, expensive to reverse, and currently being made on instinct.
Capacity, changeover, scrap, delivery dates, maintenance. Where the twelve were first written down.
Forecasting, inventory position, service levels — and the working capital locked in the wrong place.
Which site is genuinely underperforming, which is simply smaller, and where an operating plan should start.
Schedule risk that survives scrutiny, what acceleration really costs, and whether to build, buy or wait.
Claims, underwriting, clinical throughput, shared services. Queues behave the same wherever they form.
Platform and data-model decisions that commit the firm for years — and cost more to reverse than to make.
We are not sector specialists, and that is deliberate. A test that settles a capacity question on a molding floor settles it in an emergency department too. We would rather show our working in the first week than ask you to take tenure on trust.
What's slow, manual, or stuck? Send it over and we'll tell you where the leverage is. No pitch, no obligation — just a straight read on whether we can help.
Prefer email? contact@convexconsultingusa.com
