From a hand written spreadsheet to an automated, locked down investment platform

From problem to solution

1. 01 · The challenge

   The client had spent years applying a disciplined method by hand to a basket of stocks.

   Everything lived in a sheet he filled in every day: what he bought, what he sold and why. It worked, but it depended on his time and did not scale. He asked for three things: that the method run on its own, that he could test it in other markets before risking capital, and that nobody else could access its logic. And the sheet had no formulas at all: the rules existed only in his head.

2. 02 · What we did

   First we understood the method better than the sheet did.

   A custom reader rebuilt every position from open to close, we compared what the investor said he did with what he actually did and documented each rule with the data that proves it, reconciling with his own totals. On top of that we built a pure decision core, with no database or network, used by both daily operations and the simulator: what gets tested is exactly what runs. Anything that depends on the investor's judgement was not invented as a rule: the system detects it, proposes it and leaves him the final word. Above it sits a dashboard that replaces the sheet (the day's plan with the reason for each decision, portfolio, real broker executions, exportable history and a configurable basket), a market watcher that only alerts and never trades, and a backtesting lab with commissions, country specific taxes and dividends, always compared with buy and hold. Security was a first class requirement: mandatory two factor authentication, server managed sessions, a strict content policy, lockout after repeated attempts and an attack suite that runs after every change. The algorithm never leaves the client's machine.

Inside the project

Services used on this project