Forecast Monthly Demand of Automobiles to increase sales for automotive company

Project Details


Fall 2014


Fang-I Liao, Tzu Yi Lin, Po-Wei Huang , Louie Lu





Our client is an automotive corporation (e.g. Suzuki) who sell automobiles and motorcycles in Taiwan. Our forecasting goal is to forecast the demand of automobile in 2015 of four of Taiwan’s largest cities: Taipei, New Taipei, Taichung, and Kaohsiung. The forecast results will be used to set promotional plans among regions and seasons, and measure amount of automobile to import in 2015. Potential business benefits include reducing costs (inventory costs and advertisement at lower sales seasons), and improving marketing strategies (targeting the right seasons).

We consider two set of data to generate our automobile demand forecast: (1) The number of new automobile registration (2012:01 - 2014:11), and (2) The number of automobile (accumulated; 2000:01 - 2014:09). The first data help us to understand the trend in the future (upward or downward), in order to decide the long term business goal regarding increasing automobile import or the opposite. The second series is used to observed seasonal pattern, in order to make (short term) seasonal promotional plans in four regions. Since the first data has trend but no seasonality, we apply the double exponential method. On the other hand, the second data contains seasonality but no trend, so we employ the linear regression method. The validation period for the number of new automobile registration is 3 year, and for the number of automobile is an year. The forecasts of new automobile registration and the number of automobile have better RMSE than the naive forecast. The forecast results of new automobile registration indicate a steadily increase in the next 3 year, and the results of the number of automobile shows similar seasonality pattern as the historical data.

In addition, we found an opportunity for our client to boost the motorcycle sales in 2015. Since there is a policy that encourage the retirement of old motorcycles before October 2013, the number of motorcycle decline sharply in July 2013. However, since the policy is an one-time event, we expect the future demand for new motorcycle will increase. The data also suggest that the number of motorcycle regain increase in recent months.

Based on our forecasting results, we have the following 3 recommendations for our client: (1) increase automobile import in the next three year; (2) develop different promotional plans based on seasons; and (3) promote automobile sales with motorcycle sales in 2015 (e.g. discount for customers who buy automobile with motorcycle).

Application Area: