At Cencora, we handle more than 15,000 different medicines, from everyday staples like Tylenol to the latest treatments for cancer, all shipped out from 26 distribution centers across the country. To keep everything running smoothly, we need to create a separate demand forecast for each product at each location. This isn’t just important for the common drugs you see at the pharmacy, but also for special medications that might be needed suddenly. Making sure we have the right amount of everything, at the right place and time, is a tricky problem and really matters for the millions of patients who receive our drugs.
In my presentation, I’ll talk through the real-world data and technical issues we face when trying to get these forecasts right. I’ll explain what we actually mean by “demand” in our world, and share how we’ve moved past using just simple averages. Now, we use a mix of time series models and machine learning to build much more accurate forecasts for over 300,000 product and location combinations.
