[fractionalhare]

The author's thesis is that the "hard stuff" decreases over time, and gets commoditized as the field matures. My experience aligns with this view.

For example: you can theoretically do better than a seasonal ARIMA model for time series analysis. But in practice it's very difficult, and you probably don't have the amount of data you need or even an economic justification. The improvements will usually be marginal, expensive and not worth it.

Most teams developing a new product or doing research - even within large and ostensibly sophisticated tech companies - won't usefully or economically outperform something like Box-Jenkins.