Updating The Investing Engine
A year later, the tables were useful, but the assumptions were the real work.
Roughly a year after the first screens, I came back to the project with a clearer head. The early version had been useful, but it also taught me humility. Financial data is messy. APIs are moody. A missing field can look like a dead company if the code is too eager to panic.
The deeper update was conceptual. I became more interested in owner earnings: net income, depreciation and amortization, maintenance capex, working capital, and the cash that could realistically belong to owners after the business keeps itself alive. That sounds dry. It is also where a lot of the truth hides.
That led me to my own stricter little abstraction: MOG, market cap over owner earnings, divided by growth. It is simple, maybe too strict in some cases, but useful because it forces the question into the open: what am I paying for the cash engine, and what growth am I pretending to see?
The harder part is maintenance capex. That number can carry the whole argument on its back. Get it wrong and the valuation starts wearing a fake moustache. This is where the software side and the investing side meet: the model is only as good as the definitions, validation, and friction around it.
I also started thinking of investing through personas. Buffett for owner earnings and capital allocation. Lynch for simple business understanding and cycles. Munger for inversion and embarrassment prevention. Pabrai for downside and asymmetry. Damodaran for scenarios and sensitivity. The value is not agreement. The value is that they refuse to let one attractive story sit there unchallenged.
This is where I began to see the project less as a stock screener and more as a private decision-support system. Something that protects attention, forces better questions, and keeps my own enthusiasm from doing performance art with my money.