|View From The Ridge: 03|
April 3, 2015
Vice President, Services
Seasonality is present in almost everything we do and everything we are exposed to. Imagine if we dressed for the average annual temperature. Sometimes we would wear too much and other times we would not have enough. How do you think Boston would have fared if their snowfall for the past twelve months had only been based on spreading the average annual snowfall across the year? Water bills, gas bills, and even food costs can be subjected to seasonality changes. If we can accurately predict the changes that are attributed to seasonality, then we can minimize our investment in both goods and services. Ultimately, this will allow us to properly prepare for upcoming situations.
The Danger of Ignoring Seasonality
In supply chain planning it is very important to understand seasonality. You must be able to know where the natural rise and fall in demand occurs and the amplitude of the curves associated with the seasonal changes in demand. To not consider seasonality will put your business at serious risk, most notably in the form out of stocks or overstocks.
Think about the process of forecasting demand. In most forecasting algorithms there is a process to ensure that the update of forecasted demand is not overly influenced by outlier occurrences of demand. If the process is not considering seasonality, an exceptionally low or high actual demand instance might be misinterpreted as outside of what should be expected. For example, consider an item has an average rate of sale of 4, but is in a period of high seasonal demand, the normal expected demand might be 10. If this occurred using seasonality, the demand update process would recognize this as normal and provide an update accordingly. If the process was unaware of seasonality it might disregard some or all of this high level of demand. This could cause the process to understate the expected or average demand resulting in lower than expected forecasts and potentially lower than expected buy recommendations, thus causing potential service misses.
It is imperative to have an indexed process for recognizing to what extent the demand is expected to vary from the average, up or down. This will help the system to evaluate the incoming demand signals relative to what is expected.
The attention needs to be placed on exactly how to create the seasonal curves or profiles. Think about the assortment hierarchy when building profiles. Let’s assume there is a hierarchy made up of departments, classes within departments and sub classes within classes. What is the proper level to use to create the profiles? The profile needs to be built at a level that is not too high so as to hide the seasonal fluctuations of the items, but it should be high enough that data noise is not assumed to be seasonality. For example, one department might be electrical supplies, while one of the classes might be landscape lighting. It is quite possible that the seasonal curve for landscape lighting is different than the overall department of electrical supplies. A thorough and thoughtful examination of the hierarchy will help to determine the appropriate level at which to build the seasonal profiles.
Another factor to consider when building out profiles is regional influences. Think about the geographical differences in the temperature averages and temperature swings against the United States average. It would be fine if the pattern of the rise and fall was the same and just the average rate of temperature was different, but this is not the case. The curve of temperature change is very different in Miami than it is in Chicago; there is less fluctuation in Miami than Chicago. Forecasting the temperature would require a different curve. The same is true for regional demand differences for all sorts of products. Think of quadrants of the country or six regions, or a split by significant landmarks such as mountain ranges, as potential dividing lines for profile groups. If the supply chain network is very geographically focused, that might also be a good dividing mechanism.
The Bottom Line
In any supply planning solution, the forecast accuracy is of paramount importance. In order to have an accurate forecast the system needs to understand the demand signals being received. One of the most significant elements in controlling the forecast accuracy is seasonality. When the forecast is not producing the level of accuracy that you need or the resulting ordering process is not meeting your service objectives, consider seasonality. It may seem simple, but it can have the single most significant impact on the forecast quality. It is not a difficult piece to get correct, but it does need some dedicated attention.
Another area to spend time managing is lead time and lead time variability. In a later article we will look at some of the things you should consider there and how you can improve your lead time performance.
As always the folks at Blue Ridge are here to help you improve your results. Call us to help with your View from the Ridge.