AI has been used to improve supply chain and demand forecasting for a couple of decades. Demand planning applications use data driven algorithms to take historical data and use it to forecast. Demand forecasting can include promotion planning or stock and sales forecasting. The machine looks at the forecast, compares it to actual shipments, and suggests alternatives and optimization options. Over time, many companies started doing more specific forecasting for specific regions, products or stores or for more granular points in time. Both retailers and their suppliers (CPG companies or 3rd party shipping and supply chain companies) use their data to help conduct automated decision making within the supply chain.
Adding machine learning to supply chains is delivering concrete benefits for companies. Research out of McKinsey finds a large number of executives report better costs and over half report higher revenues due to the introduction of machine learning into their supply chains. Areas such as sales and demand, forecasting, spend analytics, and logistics network optimization are just part of a suite of supply chain management models.
Most demand forecasting and supply chain models require visibility into point of sale data – including inventory and sales. Many will also require production volumes, price and promotion changes as well as data around product launches. Distribution data (quantities, time, distances) is also often used for forecasting.
Demand planning applications may use a number of different types of data, such as
Competitor Pricing Data
It is important to understand the competitive environment in a specific point in time
Foot Traffic Data and Store Traffic Data
Understanding traffic according to days of the week will help anticipate the capacities needed and detect or explain sales anomalies
Sales data on different granularities regarding the specific brands and competitors can be used as forecasting targets and provide clear explainable features within the model
Any data about weather conditions is always helpful when building supply chain-related models. Pure temperature is one, but rain, snow and natural disaster conditions are even more crucial for the forecast
Additional geospatial data such as restaurant locations, store clusters, and store proximities to different points of interest could be very useful. Additionally, data about the population around the point of sale can be beneficial for clustering and forecasting purposes, and data on large scale events could also help better understand the needs in the area.
Fresh-key collaborated with DiMuto to create and implement a fresh produce cold temperature supply chain monitor.
The data logger recorded the temperature inside the shipping containers at every stage of the supply chain, and also noted how long the shipping containers spent in each location. The data monitors also tracked whether temperature values remained within acceptable margins. When the temperature was too high/too low, then the data monitors would transmit a warning.
The location data helped to visualize the efficiency of the cold supply chain, and locate areas where the shipping containers remained stationary for too long. Both parties were extremely satisfied with the test results. They are now discussing the next step in their cooperation.
Table of INEBase Trains movement. Annual. National. Rail Transport Statistics
Transport by nationality of vessel
Transport by nationality of vessel
Transport by nationality of vessel (country/regional flows from 2007 onwards)