The business case is the first prototype

Most interesting work starts with a vague mandate. Grow internationally. Do something with AI. Build a new vertical from nothing. The temptation is to start building at once, because building feels like progress.

I start with the business case instead, and I treat it as a prototype: the first working model of the idea, cheap enough to break. A spreadsheet that fails costs an afternoon. A product that fails can cost a year.

Write the decision before the model

A business case exists to support a decision, so name the decision first. Which markets open, in what order, at what price? Who signs, and what would make them say no? Everything in the model should be able to change that answer. Anything that cannot is decoration.

At Antler, the customer interviews came first, and the market sizing model and roadmap built on them decided what shipped first and what was cut. At Revolut, the new vertical I led started from a blank mandate. The business case and the five-year plan set the entry thesis, and the metrics used to greenlight and track it.

Ranges, not points

A single forecast invites a single argument: is the number right? It never is. So I build ranges. At Fever, the pricing and demand engine I wrote in Python ran Monte Carlo simulations on local purchasing power, competitor benchmarks and transaction data, so leadership committed capital against a confidence band rather than a point guess. A band changes the question from “is this right?” to “which risks can we live with?”

The same idea runs through Location Intelligence, a tool I built for choosing restaurant sites. Every zone gets unit economics with a downside and an upside case, and a zone that only works in the base case should be negotiated hard or skipped. Where a core input is estimated, the tool still shows the numbers but refuses a go or no-go verdict.

Let the case become the scorecard

At Fever, every new market started as a business case and a P&L built from scratch, and together they set which markets opened, when, and at what price. In 20 months the touring format went from 4 to 12 markets across Europe, APAC and North America. Attendance tripled, portfolio revenue grew 58%, and marketing stayed under 15% of revenue. Each new market held roughly the size of the earlier ones, which was the real test: the playbook scaled without diluting.

Approval is where weak business cases stop. A good one becomes the scorecard. I ran the weekly and monthly reviews, tracing pacing gaps back to channel, creative, price or capacity, and moving budget between markets mid-run. That turned a spiky launch funnel into steady delivery across the portfolio. Every model is wrong somewhere. When it is explicit, you can see where, and fix the plan instead of defending it.

How I build one

• Name the decision, and the person who makes it.
• Write the assumptions where a finance director can change them.
• Give every forecast a range and a downside case.
• Agree the one metric that will say it worked, before you start.
• Write down what would make you stop.

A business case that turns out exactly right is rare. One that can be wrong early, cheaply and in plain sight is the point.