Taking a cue from epidemiologists, it might be a good thing abandon the notion of "forecasts" and "estimates" in software and talk, instead, about models. A model is a dynamic decision-making aid that lets you assess the effect of various inputs on a desired outcome. 1/5

The point is to better assess the inputs, not to predict a specific output. 2/5

For example, coronovirus models are helping us keep the flow through the system low enough that we don't overload a constraint (the hospital system) by limiting inputs (people who catch the disease). 3/5

This is just solid theory-of-constraints Lean thinking à la Goldratt's "The Goal" [ https://amzn.to/2RwWkHv ]. The model is dynamic. We can play around with variables and look at the results. 4/5

This approach seems to me to be a much better assessment strategy than an estimate, which is ultimately a guess, no matter how well informed, that focuses on a fixed output (a date). 5/5

I'll add that the point of covid modeling is to maximize flow w/in the constraint, not to predict when the outbreak will be "over." That shift in focus is essential. Maximizing flow is more important to the business than delivery dates, especially if you deliver continuously.