Dabbling in Data

I worked as a software engineer for Google for 16 years. My unusual blend of C-tier coding, A-tier software engineering and S-tier data analysis combined with forcefully caring for my coworkers, the product, our users and not caring at all about career growth, propelled my career to heights that somewhat baffle me. Over the years I became jaded with the decline of Google culture (and corporate culture in general), had a daughter and decided to quasi-retire in my late 30s after being FIRE-ready for many years. I expect I will work again, but profit will not be a motivating factor. I am currently spending time with my daughter, baking, gardening, puzzling, volunteering and designing an escape room.

I started this blog as a way to clarify my own thoughts and possibly help others do the same. The name ‘Prove Myself Wrong’ comes from one of my core data tenants: it is trivial to prove oneself right, only by failing to prove oneself wrong can a hypothesis be trusted. On the surface, these are similar concepts, but one provides falsifiability and the other merely rationalization. The way I approach understanding anything is to exhaustively list what data should I expect to see and what data would show me something was incorrect. Only after I have determined these, then I look at the data and evaluate. Then I repeat, over and over again, until I am satisfied. This practice served me well in experiment analysis at Google (I oversaw the evaluation of hundreds of A/B experiments). I could find underlying bugs simply by evaluating the data, and, when others came to me with potential hypotheses, I could often immediately respond with accurate analysis from memory. This gave me an almost mystical ‘remember everything’ quality, but I consider it more akin to chess masters memory of board positions. Data always tells a story and in understanding it, it was trivial to answer questions about it. I could never remember random numbers.