Notes on data platforms, entity resolution, and the parts of analytics work that never make it into the dashboard.
I built a player dashboard to answer one scouting question, is he producing for what he costs? Here is every component, why it's drawn the way it is, and where it stops being trustworthy.
I valued every on-ball action of the 2015/16 Premier League with a standard public method. It explained the season well, except for the team that won it.
I built a weekly demand forecast that beat every standard method I tried, with almost no bias overall. I was ready to call it done. Then I split one number a different way.
A player's goals live on one site, his price on another and his shot quality on a third, and no two of them agree on his name. This is how I pulled five sources into one database the whole site can trust.
I set out to route support tickets by department. Twenty steps of honest evaluation later, the department labels turned out to be the problem, and the useful model predicted something else.