Data Science Blog

On free cash flow, every model I trained lost to a one-line seasonal baseline. FinSight serves the baseline and says so. At the end of the last post I admitted something and moved past it quickly, because it stung. For one of my three forecast targets, nothing I trained beat a one-line baseline, so I shipped the baseline. This is that story. The part that took me longest to accept is that it was not close. The most sophisticated model I built, an LSTM that won my revenue lead

The first honest backtest of my revenue model produced a forecast that looked broken in a very specific way. For companies that were growing, the prediction would track the real numbers for a while and then simply stop climbing. It flatlined, pinned to an invisible ceiling, while the actual revenue kept going up and to the right. My first instinct was that the model was undertrained, or the features were weak, or I had a bug in the reconstruction. It was none of those. The mo

A reconciliation check I had written mostly as a formality flagged Procter & Gamble's 2015 fiscal year. The four quarters I had reconstructed from their filings summed to $10,952 million of operating income. The annual figure P&G had filed for that same year was about 7% higher. Seven percent is far too large to be rounding, and far too clean to be random noise. It looked exactly like a parsing bug. I spent an afternoon convinced I had one. I did not. The quarters were right,