Inventory Optimization in E-commerce
Stochastic inventory modeling - Hagop Dippel and Andreas Syrén
About the event
Inventory optimization is one of the core problems that must be solved in the e-commerce ecosystem. To minimize inventory costs, it’s crucial to get two questions right: what quantity and when to replenish (i.e. produce, purchase, …) for a given article? Concretely, this becomes an exercise in inventory policy optimization using the forecasts of various variables.
But how do you incorporate fashion retail's high level of aleatoric uncertainty in your optimization approaches? Moreover, large applied actors have to do this for millions of articles daily.
The verdict is undeniable: your algorithms must be probabilistic and incredibly fast (exactly in the blink of an eye). We will explore one of these use cases from within a large industry player and show how you can go from probabilistic forecasts to stochastic optimization and concrete decisions.
The talk will cover:
The problem of inventory optimization in e-commerce.
A review of deterministic inventory modeling and its limitations in the face of an uncertain reality.
What types of inputs and forecasts do we need?
Using Monte Carlo simulations and gradient-free optimization, you can incorporate probabilistic forecasts in real-world inventory modeling.
How can you use Numba and cloud infrastructure to scale these techniques to tens of millions of decisions for millions of articles per day?
A small demo of practically accelerating stochastic optimization.
About the speaker:
Hagop Dippel is a Sr Applied Scientist at Zalando, focusing on building demand forecasting and inventory optimization applications. He particularly enjoys bringing research ideas to end-to-end production systems. He’s passionate about deep learning applied to real-world industry use cases and human-centred AI. Hagop studied Data Science and Econometrics at the Aix-Marseille University in France. In his free time, you can find him cycling or running (just contact him if you want to jog and/or chat!)
Andreas Syrén is a Principal Applied Scientist at Zalando. He found his love for mathematics while studying Mechatronics Engineering, pivoting into a master's in Engineering Mathematics & Computational Science at Chalmers University of Technology - before heading into industry. "I am passionate about the Applied part of Applied Science. Massive impact is out there whenever we can successfully apply scientific methodology with real-world outcomes. But for that to happen we need much more than only theory - from an engineering mindset; to people, product, and sales." In his free time, you can find him with his friends playing TTRPGs, doing some kind of boardsport, or reading up on some non-applied area of mathematics.
DataTalks.Club is the place to talk about data. Join our slack community!