Monday, November 6, 2023

Demand Forecasting (Is a Lie)

My POS company, Lightspeed, is enamored with demand forecasting. They rolled it out as a test and have now made it a permanent part of the purchasing process. I'm going to explain why demand forecasting is a mistake and why it's dangerous for a buyer to rely on it. 

Demand Forecasting is based on some basic math. If you've ever watched a TV commercial on investing, they'll add the disclaimer, "Past performance is not indicative of future results." Demand forecasting is using past performance as a guide for you to spend money for future results. If demand forecasting used AI, or if it actually used some of your past seasonal data, it may have merit. Instead, it's using some basic math to say, you should really have 6 of these, because you sold 6 in the past. 

I'm going to explain this simply, as if you were a software developer at a point of sale company, this is not how buyers buy product. Product comes in hot and cools off over time. The fact I sold a bunch of new product, is no indication I'll continue at that pace. It's almost guaranteed this won't happen. A good demand forecasting program would understand a trend in sales.

Demand forecasting is also not supply forecasting. The system has no idea how to calculate these variables:

  • Seasonal Stock. Am I going into or coming out of a busy season? A good demand forecasting system could look back over a year and perhaps know. 
  • Safety Stock. The Fantasy Flight Games solution to out of stock situations is to always have more on hand. A POS system doesn't know what stock is safety stock. This is stock to have on hand to avoid outages, because reasons.
  • Flash Stock. Much of our sales in this trade includes items that come out hot and peter down to slower sales very quickly, like collectible card games. Understanding that I will not have sales levels of days 1-30 of a Magic release in days 31-60 should be part of demand forecasting.
  • Space. There is often no consideration for space available in demand forecasting. My demand forecasting POS suggested I order 50 cases of Mexican Coke, based on past sales. Where the hell am I going to put 50 cases of Coke? Getting it weekly makes a lot more sense than a 30 day supply.
  • Budget. Of course, the biggest problem with demand forecasting is it will suggest you obtain an optimal stock level, without knowing your budget. It doesn't integrate with an Open to Buy spreadsheet. It doesn't know you're saving up for a big release.
So did I try demand forecasting? I did! I realized demand forecasting is a perfect representation of what I should have ordered a month ago. Demand forecasting is a look back, a score card for your previous purchasing. Until it has artificial intelligence involved, where it can look at larger periods of your sales, and perhaps work its way into the Internet to provide intelligence, it's really more of a gimmick. If it had AI, it would be amazing. Imagine a system that could do that! Instead we get some basic formulas using limited data that look backwards instead of forwards.

Don't get suckered into demand forecasting. It would be like using a self driving car that only had cameras in the back. If you're using Lightspeed, I suggest you open a ticket and tell them to get rid of this clutter.