The following table contains quarterly data on Upper Midwest car sales (CS) in thousands for 1996 Q1 through 2016 Q4:

                        Upper Midwest Car Sales (CS)

Year	Q1	Q2	Q3	Q4
1996	407.6	431.5	441.6	306.2
1997	328.7	381.3	422.6	369.4
1998	456.3	624.3	557.5	436.7
1999	485.0	564.3	538.3	412.5
2000	555.0	682.7	581.3	509.7
2001	662.7	591.1	616.9	529.7
2002	641.2	632.7	576.6	475.0
2003	542.8	558.9	581.7	537.8
2004	588.1	626.5	590.9	580.1
2005	589.2	643.2	593.9	612.2
2006	586.1	699.4	734.4	753.8
2007	691.6	793.4	864.9	840.8
2008	653.9	754.8	883.6	797.7
2009	722.2	788.6	769.9	725.5
2010	629.3	738.6	732.0	598.8
2011	603.9	653.6	606.1	539.7
2012	461.3	548.0	548.4	480.4
2013	476.6	528.2	480.4	452.6
2014	407.2	498.5	474.3	403.7
2015	418.6	470.2	470.7	375.7
2016	371.1	425.5	397.3	313.5

a. Prepare a time-series plot of Upper Midwest car sales from 1996 Q1 through 2016 Q4.

b. Use Forecast X™ to do a time-series decomposition forecast for 2017 (be sure to request the MAPE). In the results, you see the seasonal indices. Do they make sense? Why or why not?

c. Forecast X™ calculated the historic MAPE as a measure of fit. Write a short explanation of what this MAPE means to a manager. 

d. Now calculate the MAPE for the 2017 Q1–2017 Q4 forecast horizon as a measure of accuracy, given that the actual values of CS for 2017 were:

2017 Q1	301.1
2017 Q2	336.7
2017 Q3	341.8
2017 Q4	293.5

e. Prepare a Winters’ exponential smoothing forecast of CS using data from 1996 Q1 through 2016 Q4 as the basis for a forecast of 2017 Q1–2017 Q4. Compare these results in terms of fit and accuracy with the results from the time-series decomposition forecast.