#26 Software Estimation Lessons Learned from Covid-19 Forecasting

Inspect and Adapt

For the past year, Steve McConnell has applied his extensive estimation expertise to a timely problem: Covid-19 forecasting. Steve’s Covid Complete Data Center provides US national data, state data for every state, state scorecards, forecasts, forecast evaluations, and other data on the pandemic: https://stevemcconnell.com/covidcomplete/ His Covid Complete forecasting model has been accepted into the US Center for Disease Control’s “Ensemble” model, which means that it is one of the models driving overall CDC forecasting. In this episode, host Mark Griffin and Steve explore what Steve has learned from his modeling efforts and the lessons learned that are valuable for the software world. You’ll learn the importance of the following for software estimation: using historical data, keeping "control knobs" to a minimum, the difference between accuracy and precision, the difference between reported and actual ground truth, and the absolute necessity of closing the loop and judging your forecasts’ accuracy and effectiveness.