Guest Post: Forecasting Lessons Using PC Sales

Prof. Howard Weiss presents an interesting, real-world example of seasonality and forecasting.

If we examine PC unit shipments in the U.S. 2013-2023, by quarter, there are a couple of interesting lessons we can learn from the data.

The data are separated into pre-Covid and Covid time periods because it is obvious that the graph looks different before 2020 than at 2020 and beyond. If you look closely at the pre-Covid data, it is very easy to see the seasonality. Quarter 2 is higher than quarter 1, Quarter 3 is higher than Quarter 2 and Quarter 4 is higher than Quarter 3 in EVERY year from 2013 to 2019.

Chapter 4 of your Heizer/Render/Munson textbook discusses Seasonal Variation in Data. Using Excel OM for the method of Example 9 we find that the seasonal indices are as given in the table below for the pre-Covid period. In addition, using regression we find the line that fits the data best is:

Shipments (in millions) = 15.675 – .06*x

where x is the time period from 1 to 28.

Notice that shipments have been decreasing by 60,000 units per year. Using the regression equation, the forecasts for the next 4 periods in 2020 are given in the table

Pre-COVID Percent of Demand Seasonal Factors X value

(2020)

Forecasts

15.675 – .06*x

Actual (2020 data)
Quarter 1 21.6% .862 29 13.935 10.83
Quarter 2 25.1% 1.003 30 13.875 15.70
Quarter 3 26.4% 1.057 31 13.815 23.62
Quarter 4 26.9% 1.076 32 13.755 19.03
Total 55.38 69.28

 

Looking at the actual 2020 data, it is obvious that Covid caused a significant increase in PC shipments. The increase is even more pronounced in 2021. This is not surprising as more and more students and workers were working from home rather than in the office or university. Also, examining the graph, the seasonality for 2020-2023 is not as obvious as for the pre-Covid period.

During COVID Percent of Demand Seasonal Factors
Quarter 1 22.3% .893
Quarter 2 26.3% 1.052
Quarter 3 26.8% 1.072
Quarter 4 24.5% .982

 

When forecasts for 2020 were made in 2019 it was impossible to know that Covid would strike and affect shipments as much as it did. But by quarter 2 of 2020 it was clear that quantitative forecasts based on past shipments would have large errors. At this point it would be imperative to introduce a qualitative method into the forecasting process, as discussed in the chapter.