For consumer packaged goods companies serving retailers, forecasting the true demand for perishable products is the key to maximizing revenue while reducing the costs associated with salvaged goods. This can translate into millions of dollars of savings and additional revenue annually for some companies.
Through the application of machine learning, artificial intelligence (AI), and big data analysis, companies can take advantage of growth opportunities and cut their salvaged goods costs by as much as one-fourth, especially with perishable goods where there’s a persistent tradeoff between an oversupply of goods and missing out on sales by not providing sufficient inventory.
Improved forecast accuracy can be attained through reinforcement learning, a machine learning technique that helps determine which actions will lead to the greatest rewards. Combining the potential of reinforcement learning with human decision making can deliver far more effective forecasting models, for both direct to store and centralized distribution business frameworks.
Multi-arm bandit algorithm
At the core of reinforcement learning is something that statisticians call the “multi-arm bandit” algorithm. It refers to the analysis of a group of multiple slot machines (one-armed bandits) in a casino, and the testing and learning involved in determining which of those machines will pay off better than others in the long run. Once determined, a gambler can then focus his or her time and money on those machines for a better return. Learn how this algorithm applies to making more advantageous forecasts for perishable products.
Most companies look at demand from a historical perspective. They stock the shelves based on traditional, seasonal and promotional patterns. With the science of reinforcement learning, forecasts become more accurate, allowing companies to take advantage of revenue opportunities in rapidly growing geographic areas or in new markets.
In the process of leveraging a growth opportunity, a supplier doesn’t want to take too great a risk of increasing wastage. Doubling inventory on a shelf covers most demand scenarios, but also boosts wastage. Conversely, empty shelves take care of the wastage problem, but reflect an unknown number of missed sales opportunities. With AI and machine learning, a company uses advanced calculations versus historical data alone.
Building a better forecast
The learning process involves constant projections of how much more, or less, bread or milk or fresh vegetables should be placed into retail stores, and when, and taking lessons from the results. The ever-smarter system will tell you, essentially, “Here is where you have the best growth opportunity and least chances of return for this product.”
Reinforcement learning does not replace the human element; rather it supplements human insight. People make the decisions, only they do it with far better data analysis to back them up