Leveraging Big Data Engineering for Predictive Analytics in Wholesale Product Logistics

Abstract

Unlocking the latent insights of Big Data has been a holy grail of predictive analytics since the term Big Data was coined. Yet, it is only very recently, on the back of technological advancements, with scalable data engineering tools, an ecosystem of data lakes, massively parallel processing analytical databases, predictive modelling libraries and, of course, the rise of cloud computing, that this goal has become achievable. What had always been a herculean task of extracting value using predictive modelling, has, thanks to these Big Data-ready tools, transitioned towards an almost cookie-cutter capable endeavour. Delivering production-level models at arbitrary workloads and presenting the options to individual stakeholders has now become possible for an organisation's sizable analytical support team. This paper intends to showcase one such use case within the landscape of wholesale product logistics: at large scale, automating demand forecasting with predefined intelligence - a mixture of pre-learned or commonly known pieces of information – for the creation of demand plans of a considerable product segmentation for a large customer base. Delivering daily production forecasts at daily warehouse replenishment levels, monitoring and controlling variability of predictive performance to continue to administer a key influence area of much of logistics execution. The objective being the increase of operational efficiency, improvement of service levels with data-driven decision-making, and even influencing tangible aspects outside the enterprise when creating a plan that balances demand against partner replenishment capabilities. It uses the previously mentioned tools to solve the need of future demand planning through volume indicators which serve to take replenishment actions, considering product storage capacities at different partner companies: those are wholesalers, retailers and distributors.