We run a sales forecasting model every month to try to predict how we’ll do for bookings one and two months out. Our model is a neural network model based on company performance, econometric data, and market trends. This type of model uses years of historical information to try to develop predictive trends. We run the model multiple times to get a range of predictions. Sometimes they are tightly grouped, and other times, they can bounce around. This happens even though the same data are used. When things bounce around, there are usually conflicting data. Data that usually move together begin to come apart. That’s what happened in December. Good econometric information was clashing with negative emotional measures and the predictions bounced back and forth between dismal and pretty good. The real, numerical data was showing growth, but people were still negative on the economy.
We were expecting that the emotions would begin to turn to mirror the numbers, and that began to happen in January. Prior to January, most of of our bookings had been from recently added customers, seemingly unaffected by the slow economy, while our legacy customers, more affected by the economy, were not ordering. That changed in January. We began to see orders from companies who had been dormant for the past few years. Additionally, the newer customers started ordering larger quantities of parts. This is beginning to feel more “normal”, if such a term can describe our post recession economy. In any event, we’re enjoying the change. Hopefully, it will continue. Maybe the forecasting model will start showing more stable numbers and operate as if the world has returned to a more predictive mode.