Profitability you can prove, not guess.
An operating model of your business showing which customers, products, and activities drive your profit, and where it leaks. A profitability engine does the calculation; and the AI you already use can now drive it. Run by your finance team.
Watch a worked exampleMore revenue isn't more profit.Your accounts tell you the company made money. They don't tell you which work made it, or which work quietly cost it.
Most operationally complex businesses run on a P&L that aggregates everything above the gross margin line, then aggregates everything below it. It tells you whether last quarter was good or bad. It doesn't tell you which customers were quietly subsidising others, which products earned their place on the shelf, or which operational activities consumed margin you can't see. Attributing cost properly answers all of it, and that is exactly the work most teams skip because it has always been too slow and too expensive to do well.
The people who feel this most acutely are the ones closest to the numbers. Finance leaders who close the books cleanly every month, then get asked by a CEO or a board which customer segments to double down on, and want to answer with evidence, not instinct.
Curious what your profit shape looks like?
Answer five quick questions and see the likely shape of your profit: where it concentrates, how many customers may be losing you money, and the margin probably at risk. An illustrative benchmark for businesses like yours, not a guess.
How it works
Map your operations.
A guided process turns your business into a structured operating model: revenue streams, activities, cost pools, products, customers. Days of work, not months.
Connect your data.
Pull in financials from your accounting system and operational data from wherever it lives. ERP, CRM, spreadsheets, warehouse systems. We work with the data you already have.
Let the engine do the work, and the AI do the analysis.
CostCtrl's allocation engine attributes every cost to the activities that consumed it, and every activity to the customers and products that triggered it. Tens of thousands of consistent, explainable allocations. The AI reads the result and tells you what to do with it: which customers to renegotiate, which products quietly lose money, where margin is leaking, and what a change to price or mix would do. The analysis a team would take weeks to assemble, ready the moment the model runs.
Act on what you find.
Profitability by customer, product, region, channel, and activity, with scenario modelling for pricing, mix, and operations. Decisions you can make with confidence, margin you can win back, and numbers your finance team can stand behind in a board meeting.
Then close the loop.
The model above is deterministic and auditable. The MCP server adds a port to it, so the AI you already use can drive it: set a target and constraints and let it search for the plan. Goal seek, for the whole business. Every candidate computed by the engine, not guessed.
- EngineCostCtrl
The deterministic operating model. Cost pools, drivers, attribution, every rule auditable.
- Diagnostic portMCP server
A standard socket. Scenarios, inputs, calculations and results, exposed as tools any AI can call.
- Diagnostic computerThe AI you already use
Claude, ChatGPT, Copilot or your own. It reads, proposes, changes inputs and asks the engine what happens.
The agent never types a number into the P&L. It changes inputs. The engine computes the consequences, and every figure carries the rule that produced it.
Get net margin to 7% in the FY2027 plan. Lose no more than 3% of revenue. No headcount reductions. No plant or hub closures. Find the plan.
What the engine computed- Revenue down 1.6%, inside the 3% limit
- 526 people in, 526 out
- 10 scenarios computed, 2 rejected by constraints
Illustrative. The demo manufacturer's figures, computed by the CostCtrl engine.
What you can see, the day after you finish modelling.
The current options for understanding profitability all break in the same way.
Spreadsheets.
An analyst builds a model. It works for a quarter. Then someone leaves, the data shape changes, or a formula breaks. You're back to instinct.
Enterprise platforms.
Powerful, capable, and built for companies that can afford a six-figure implementation, an in-house consultant team, and a year before anything ships.
Consulting projects.
A team of smart people produce a profitability deck. The deck is brilliant. By the time you've acted on it, the underlying business has moved on.
Ask ChatGPT on its own.
A general model answers any what-if fluently and instantly. It has no engine to compute against, so the figure is invented, and nothing you would put in front of a board.
CostCtrl is the option in between.
A hybrid that takes the best of each without the downsides: the rigour of the enterprise platforms, the speed of a consulting project, the ongoing usability of a tool your finance team actually owns, and an engine for the AI to compute against.
See what CostCtrl finds in your business.
Book a 30-minute call. We'll walk through your questions and send a tailored proposal within 48 hours.
Book a 30-minute call →How Snell turned a hidden cost-to-serve problem into a multi-year margin turnaround.
CostCtrl gave us a view of our business we'd never had: which customers and products were actually contributing, once you accounted for everything it took to serve them. That visibility changed how we price, what we focus on, and how we work with customers. It's become part of how we run the business.
negative contribution cut from $1.335M to $665K
from 830 down to 295, most near break-even
a model Snell owns and runs, refined as the business changes
Driving EBITDA across a portfolio?
CostCtrl is used by operating partners to deploy profitability programs across portfolio companies. Repeatable methodology, deployed in weeks, transferable across operationally complex industries.
Common questions.
How does it go from start to live?
We start with a Forward POC: an engineer-led proof of concept, built fast with our own AI tooling, that shows CostCtrl running on your real data and questions in a couple of weeks. If it earns its place, that same work becomes your live operating model with no restart and no second build. From there your finance team runs it day to day, with as much or as little support from us as you want.
Do we need to connect our ERP?
No. We pull from the systems you have, in the formats they produce. Most engagements start with monthly data exports and add automated connections later if useful.
Where does our data live?
In a secure, logically isolated environment on AWS in the EU. CostCtrl never uses customer data to train AI models. The specifics are on our security page.
How is this different from our BI tool?
A BI tool reports on the data you already have. CostCtrl enriches it first: our engine attributes every cost down to the customer, product, and activity that drove it, so you see profit at a granularity your source systems never captured. A dashboard shows you what happened. CostCtrl tells you what it cost, and what to do about it.
How is the AI used, and is it reliable?
The work is done by a proprietary, deterministic allocation engine we have built and refined over years: it applies consistent rules to attribute costs across every customer, product, and activity. The AI sits on top and turns those results into insights, visualisations, and plain-language analysis. The engine computes; the AI explains, and through the MCP port it can also drive the model toward a target. In every case the engine computes and the AI never supplies a figure. Every attribution is explainable and auditable, and nothing is generated, only computed, which is what makes it reliable in a way a general-purpose AI on its own never could be.
Can we plug our own AI into it?
Yes. CostCtrl can expose your model through an MCP server, a standard port that Claude, ChatGPT, Copilot or your own AI can connect to. The agent can read results, create scenarios, change inputs and run the calculation. The engine computes every consequence; the agent never types a number into the P&L. Opening the port and choosing the provider is your decision, made under your own agreement with that provider. Our security page sets out how that works.
What does it cost?
Most customers pay between $1,500 and $3,000 per month (USD), on an annual agreement billed monthly. A distributor carrying around 3,000 products on $30M in revenue typically sits mid-range; where you land depends on the complexity of your operating model and the compute it requires. Most customers see it pay back many times over in margin recovered and sharper pricing. We'll send a tailored proposal within 48 hours of the call, once we understand your specific situation.
What if we already use a different cost platform?
Some customers run CostCtrl alongside an existing platform; others switch over. We're happy to talk you through the trade-offs honestly.
See what's hiding in your operations.
Book a 30-minute call with Sam or Miguel. We'll walk through how CostCtrl would map onto your business, with one of your real questions in mind.