I am sure you are very familiar with the shopping extravaganza during the holiday season with all the promotions, discounts and attractive prices. It is predictable to the extent that companies rely on past experience to ensure the right amount of inventory is available where and when needed. But this requires more than just a statistical analysis and regression formulation to predict expected demand. It has a lot to do with social behavior of the buyers and how they react to changing prices as well as many other factors. Do they buy a month in advance or 1 day in advance? Are they waiting for more price drops until the last minute or they just want to buy and get it over with? What are the best days to have promotions, how many days in advance of Christmas day? On a week day or weekend? What is the role of the weather in customer behavior? How is it different in low income residential areas vs. higher income areas and are they inclined to buy on-line or walk into a store? There are plenty of other factors that we call “attributes” that are at play here and are not captured in a simple statistical analysis.
Applying machine learning techniques however can lend much more accurate results. ML can deal with hundreds of different influencing factors and conclude relationships that may be hidden to us and their relevance on the outcome that we are looking for. Humans do this all the time. A simple example is if we have a headache a pain killer generally helps. This is based on past observations and experiences. However, the cause we may not be aware of. It could be air pollution. The latter is not obvious to most of us. With enough data, a ML algorithm can relate the two as cause and effect. Traditional demand planning solutions are very limited in terms of the dimensions that they consider. Using ML technology, the limits are almost endless. Happy shopping and happy holidays to all.