Merriwell Bag Company - Analysis of Quantitative Forecasting Models

Background

Merriwell Bag Company (Merriwell) is a small, family-owned corporation with stock equally divided amongst five members of the family. Ed Merriwell, the founder of the company, formed the business over 20 years ago and originally sold bags to a small discount store and a regional chain of drug stores. As these two customers grew so did Merriwell, and today these two original clients are Merriwell's largest customers. The company has found a market niche by supplying stock bags to various small chain stores scattered over a wide geographical area.

Merriwell currently has over 500 customers and they do not aggressively pursue more. The Merriwell family is cautious about becoming too heavily reliant on any one customer; therefore they have the policy that no single customer can account for over 15 percent of sales.

Merriwell only manufactures pinch-bottom general merchandise bags. They have a central strategy built around low unit cost production through standardization. This allows them to have a competitive selling price with large bag manufacturers. Merriwell takes great pride in "taking care of" a customer who has an emergency need for additional bags or who would like Merriwell to warehouse a bag order for a given time because of storage problems at the customers' warehouse.

Forecasting Issues

Providing such a personal service requires Merriwell to have tight inventory control and production scheduling at the bag plant. A highly accurate demand forecast allows Merriwell to service special customer requests by use of Merriwell's own warehouse facilities and routing schedules of the company's truck line. Due to the constant growth of accounts and changes in personnel in customer purchasing departments, the accuracy of Merriwell's forecasting has been declining rapidly. The percentage of short-shipped accounts for particular types of bags is drastically increasing and the warehouse is becoming overstocked with other types of bags.

The demand forecasting for the bags has always been difficult to predict due to the seasonal fluctuations in sales. There is usually an increase in demand for these bags prior to the holiday season. Therefore the Merriwell family needs a forecasting method that will take this seasonal factor into consideration. They also require a forecasting method that exhibits stability and anticipates the growth patterns of their respective customers.

Analysis of Quantitative Forecasting Models

Merriwell should incorporate qualitative and quantitative methods into its forecasting of sales. They should utilize a qualitative method because it utilizes managerial judgment, experience, expert judgment, relevant data, and an implicit mathematical model. A qualitative method will be based on educated opinions from appropriate people and may be able to signal toward events that a quantitative method may not capture.

For the purposes of accurately determining sales, Merriwell should incorporate a quantitative method of forecasting. The appropriate quantitative forecasting method for Merriwell's intended use is a time series method. Time-series methods are used to make detailed analyses of past demand patterns over time and project those patterns forward into the future. These forecasting methods are based on analysis of historical data and make the assumption that past patterns in data can be used to forecast future data. Another assumption of time-series methods is that demand can be decomposed into components such as average level, trend, seasonality, cycle, and error.

In determining the quantitative forecasting model that will best work for Merriwell, we decided to look at their monthly sales over the last five years and determine which forecasting model would have most accurately predicted the actual sales over that period of time. We used the following data in creating the quantitative forecasting models:

Table 1.1 Monthly Sales 2003-2004

Sales (in number of bales)

Month 2003 2004 2005 2006 2007

January 2000 3000 2000 5000 5000

February 3000 4000 5000 4000 2000

March 3000 3000 5000 4000 3000

April 3000 5000 3000 2000 2000

May 4000 5000 4000 5000 7000

June 6000 8000 6000 7000 6000

July 7000 3000 7000 10000 8000

August 6000 8000 10000 14000 10000

September 10000 12000 15000 16000 20000

October 12000 12000 15000 16000 20000

November 14000 16000 18000 20000 22000

December 8000 10000 8000 12000 8000

Average Forecasting Model

We first looked at the average forecasting model. This model accounts for the average of sales over the past five years (See Attachment A). The model predicts that sales will remain constant for the next year and it does not account for the seasonality and growth of sales on a monthly basis. For these reasons it appears this model is inappropriate for the forecasting needs of Merriwell because it does not accurately predict the actual sales for a given month.

Moving Average Forecasting Model

2-Month Moving Average Forecast

The next forecasting model we attempted to look at was the moving average forecasting method. This forecasting method assumes that the time series has only one level component plus a random component. It does not take seasonal patterns, trends, or cycle components into account because they are assumed to be present in the demand data (See Attachment B).

We first looked at an annual moving average forecast which looked at total annual sales to obtain the moving forecast. It is clear from the graph that the 1-year moving average is slower to respond to demand changes than the 2-month moving average forecast. The 2-month moving average forecast more closely resembles actual sales but because it doesn't account for seasonal patterns or trends. For these reasons the moving average forecasting model is unable to accurately predict Merriwell's sales for a given month. As seen on the graphs above, this forecasting method is not appropriate for Merriwell's forecasting needs.

Exponential