CHAPTER ONE
Measure
The farm and the workshop
I grew up on a coffee farm in Africa. Nobody there called it probabilistic reasoning, but that's what it was: when to plant, when to pick, what the rains would do, what the price would be by the time the harvest reached market. Every season was a portfolio of decisions made under uncertainty, with real consequences and no dashboard. I watched people get good at it — not by eliminating the uncertainty, but by developing judgment about when they knew enough to act.
At nine I became an apprentice carpenter. Furniture was the thing I loved — building something from raw wood that would live in someone's home. Carpentry has an unforgiving honesty about it: a chair isn't good because the joinery is clever. It's good because someone sits in it every day for years. The craft is invisible when it works; the only measure that matters is use.
Those two lessons — act well under uncertainty, and build for daily use — turned out to be my entire career. I just didn't know it yet.
CHAPTER TWO
Learn
The reader's detour
The third thread was books. I read philosophy from an early age, and it lit a different fire: the ideas underneath a thing should be true, not just the surfaces. That pull took me back to school at nineteen — a computer science degree, then project management, then an MBA.
The carpenter and the reader have never left. At McKinsey, the reader was honored twice with the Gluck Knowledge Award — the firm's recognition for advancing what it knows, not just what it sells. I mention it because it marks the standard I hold the work to: rigor underneath, or the build doesn't deserve to stand.
CHAPTER THREE
Cut
Where the measuring meets the decision
Every product I've shipped since is the same product, in different clothes: a probabilistic system on one side, a person who has to decide on the other.
- $2.3B+ / yr Fraud models inside a mobile wallet moving billions across ten countries — worthless until the operators trusted them enough to act.
- $10B+ · 5,000 reps Pricing products at McKinsey guiding transactions worldwide — a self-learning system that sales reps actually followed, worth 1.5–5 points of margin because they did.
- 10 → 60 markets A Bayesian marketing-mix platform for a global beverage leader — where planners allocate real budgets through an optimizer that shows its uncertainty instead of hiding it.
The pattern never changes. The model is the measuring. The decision is the cut. And the product is whatever it takes for a person to trust the first enough to make the second.
Have a model people don't act on? That's not a modeling problem.
Let's talk.