Optimizing workforce planning by incorporating uncertainty forecasting

Sales forecasts are predictions and therefore always come with uncertainty. Biases in the forecasts result in plannings mismatching demand and subsequently resulting in either costs of over- or under capacity. However, the costs of mismatched capacity are not equally expensive and therefore form an optimization problem. A classic and often referred example of this is the newsvendor optimizing the amount of papers to purchase given its forecast uncertainty. Although those cases might seem very similar to our optimization problem, our case differed in a way that days are not independent of each other, under capacity results in postponement of work the next day including financial implications. In this project we were able to develop a solution by combining the usage of monte carlo simulations and a simple non linear optimization model. In a Streamlit web application we provided stakeholders the opportunity to play around by letting them modify capacity constraints and running simulations to generate costs projection.