The Unusual Part Is the Combination

What I Actually Do

I build the thing that produces the decision, not the report that describes the past. That means a pricing model that says which accounts you can raise and by how much, a forecast that states its own error band and what that error costs you, a mix model that tells you what to run when the factory is full. The output is a recommendation with the arithmetic attached, so it survives the questions it will get.

What I Have Built, and You Can Check

Everything below is on this site, free, with no signup. It is the fastest way to judge whether the thinking is any good.

6 formula-driven Excel models, covering budgeting and variance, price/volume/mix decomposition, driver-based forecasting with sensitivity ranking, a 26-week cash stress test, a Solver optimiser with 5 decision variables and 6 business constraints, and a manufacturing planner built on backlog and capacity.

6 browser calculators, each settling exactly 1 decision: the true contribution of a customer after cost to serve, the volume a discount has to buy back, the price your differentiation actually justifies, the mix to run against a binding constraint, which variances are signal, and whether a sharper forecast would change anything.

3 packaged toolkits, each running a whole recurring job end to end: the monthly close from raw financials to a board pack, a driver-based forecast of where the year lands, and a pricing and mix decision set with the arithmetic attached. 15 published working notes, each carrying the method and the numbers.

Where the FP&A Comes From

Corporate FP&A in a business with genuinely awkward drivers: pricing dynamics, product and customer mix, gross margin pressure, capacity constraints, backlog, and working capital. Budgeting, forecasting, variance analysis, management reporting, margin analysis, and cash-flow planning, on a monthly clock.

That is where I learned the part no degree teaches: the data is incomplete, the deadline is real, and leadership needs a defensible answer this week rather than a perfect one next quarter. Every model I build is shaped by that constraint. If a team cannot run a model without its author, the model has failed.

Where the Quantitative Depth Comes From

Graduate training in economics and econometrics, which supplies the tools that rarely show up in standard FP&A: estimating the relationship between 2 business variables rather than assuming it, separating price from volume from mix, deciding whether a movement in the numbers is a signal or noise, and quantifying uncertainty in a form a decision can use instead of a caveat nobody reads.

In practice that is demand and revenue-driver analysis, price sensitivity and elasticity, regression-based modeling, and scenario work that moves the linked drivers together instead of nudging 1 input at a time.

Optimization, and Why a P&L Cares

Mathematical programming, applied to pricing, mix, and resource allocation. The distinction matters commercially: a scenario tool tells you what happens under the 3 options someone thought of, an optimiser searches the whole feasible set and tells you which constraint is costing you money. On a full factory that is often the difference between a strong top line and a falling profit.

The background runs to large-scale nonlinear optimization modeling in GAMS for policy scenario analysis. I mention it because it explains the depth behind the business tools, not because it is a separate practice.

How I Use AI, and How I Know

I benchmarked it rather than guessed. 2 models, 5 conditions, 7 scoring dimensions, on a real CFO monthly close task, with the research paper published alongside the results and 1 clear risk warning attached.

The conclusion I work from: AI takes the assembly and the first draft, and gives the hours back to interpretation. It does not do the interpretation, and treating the 2 as interchangeable is the single most expensive mistake available in finance right now. The workflows I build put the boundary in writing.