I work as a Materials Management Analyst for Portland General Electric (PGE) – the electric utility company that serves the Portland, Oregon metro area. It goes without saying how critical it is that our organization prepares itself for the unknown so that our customers can have electricity every single day. To this end, I use forecasting to help understand what our material needs will look like.
Forecasting For A Massive Number of SKUs
In the electric utility world we are constantly consuming materials that make up our transmission and distribution infrastructure. You can imagine what these materials are if you look up at a telephone pole in your city. These parts, pieces, widgets, and hardware make up a material catalogue of over 6,000 different parts!
Because these materials are constantly being ordered in and consumed out of our inventory it is essential that our organization focuses on executing the right material, at the right quantity, at the right place, at the right time. What exactly is right for each material?
In order to prepare ourselves with the correct quantities of materials I create forecasts to help align our future demand with our supply base. At PGE, we use Oracle as an ERP which provides us with a robust and sophisticated system to report on usage history, by material, by location, by month. With this historical usage data, we can easily build out forecasts that help us understand what usage in the future will look like and answer that question of what is the right quantity for each material.
The Forecasting Models I Use
I use several different forecasting methodologies to help understand our future demand. I keep in mind that different forecasting methods can better serve different forecasting ends. For example, if I were considering what we will be needing one month from now, I may be better off using a Weighted Average model or a Moving Average model.
These models can be better predictors of the near-term because they give more credence to the recent past. If I were considering our demand five years from now, I will defer to a Linear Regression model in which the model attempts to predict the future by regressing to the mean of the data sample. The Linear Regression model will not attempt to hug the curves of the data, so to say, the same way that the other models will but it will reduce the risk of the forecast being wildly off in any one month.
The Error Metrics I Use
For the vast majority of our materials that we forecast, the lead-times are an average of 45 days. For these materials I maintain a rolling 12 month forecast horizon and I find that the Weighted Average model is adequate for predicting future demand. This is also born out in the metrics that I use to assess the performance of the forecast. I rely on metrics such as the Mean Absolute Deviation or MAD (AKA Mean Absolute Error) as well as the Mean Absolute Percentage Error or MAPE. In short, these metrics can show how close the forecast was to the actual demand.
Managing Outliers/Extreme Events
The Weighted Average model also has the benefit of allowing me to change the weights I apply to each year. For example, here in the Pacific northwest, we experienced some remarkable weather in both the years 2021 and 2020. In 2021 the Portland Metro area endured what has become known as the, “Great Ice Storm,” an ice storm that resulted in unprecedented amounts of customers being impacted as well as material usage.
In 2020 much of the Pacific northwest faced forest fires that also made for remarkable material usage. If we were to consider the usage history in these years, our forecasts would likely be inflated or overstated. The Weighted Average methodology allows me to take these incredible years into consideration and place a lower weight on the model. This helps keep the results of the forecast more aligned with previous years that have lower Standard Deviations of consumption history.
As explained above, the forecasting helps, ultimately, align our consumption with our supply base. PGE is a vertically integrated company that relies on the partnership of its suppliers to maintain the materials needed to carry out PGE’s mission. Providing forecast information to our suppliers, especially to those suppliers providing materials with long lead times, puts them in better position to execute our orders. The result of this collaboration and partnership is that our suppliers know what kind of inventory to carry for PGE to meet our monthly consumption rates and what kind of safety stock levels to carry.
We will never be perfectly prepared for what the future will bring to us and what kind of demand we may encounter but forecasting and a proactive approach allows us to be agile and quick to respond when unforeseen spikes in demand occur. Forecasting is imperative to our operation and we must be able to continuously improve the efforts to ensure that we are prepared for the next big event, whatever that may be.
