Inventory management in times of environmental uncertainty

In the decade since the credit crisis, businesses have been adapting
to a new reality. Uncertainty has increased in areas that were long
considered stable.Until recently, firms doing business in Britain were
mainly concerned with making strategic decisions with little to no
information on fundamental issues such as the future customs
regime. Similarly, American companies were defining business
strategies during an ongoing and uncertain trade war with China.
Then, at the start of 2020 the covid19-crisis exploded and the world
changed overnight. Brexit, trade wars, and covid19 introduced
different shocks to the worlds' supply chains. From a modeling
perspective, these shocks represent non-stationary uncertainty. Their
impact is hard to predict (hence, uncertain) and they change the
underlying structure of the system (hence, non-stationary). Such
newfound uncertainty requires a new decision-making paradigm.
Statistical models, using past data to predict future behavior, are
powerless under such conditions. In this project, we study how
inventory decisions can be adapted to accommodate this new reality.
We propose a three-stage approach: understand, model, refine. First,
we use the covid19 shock as a natural experiment to understand its
effect on firms and industries. Then, we use machine learning
techniques to develop inventory models under non-stationary
uncertainty. Finally, we draw on our domain knowledge to propose
methodological refinements for solving such models.