A new approach to calculate safety stock in the semiconductor Solid-State Drive (SSD) supply-chain.

Conventional models for demand forecasting assume that market demand at each time stamp, t, is normally distributed. This assumption is often baked into commercial software for supply planning and cannot be modified. But, upon studying a number of historical SSD programs, we observe that demand is not normally distributed in Solidigm’s SSD supply chain and instead demonstrates varying behaviors at each time stamp, t, across the manufacturing and order fulfillment timeline. I propose an alternative algorithm that leverages non-parametric kernel density estimations and overlapping continuous time intervals to optimize safety stock and inventory levels.

I presented a research poster on my findings at the Institute for Operations Research and Management Science (INFORMS) Annual Meeting in 2022.