Cc Warehouse

CC Warehouse Report

To assist Ms. Neha Joshi, the manager of CCW on the project called “Call volume forecasting”, the following methods can be used to forecast the volume in the subsequent quarters:

Naïve method:

This uses the last value as the forecasted value and is highly risky because it does not consider the trend or fluctuations caused by the seasonal patterns in demand. The method is very simple but poses a huge risk in terms of forecasting:

[pic 1]

There is a steep fall of demand which possess a high risk of error for this method.

Moving Average Method:

A specific time frame is taken in this case instead of the last value. Considering the number of periods to be considered (i.e. n = 3) in this case we have predicted the future points:

[pic 2]

We can see that the errors are minimized as compared to the naïve method but subsequently if we take more number of points instead of 3 the errors can be bigger since the environment will change due to prolonged period of service.

Weighted Moving Average:

In this method, specific weights are associated to different points in a particular time frame accordint to its relevance. The sum of the weights must equal to 1. We have to take extra precaution in case of chance events in this case. If due to a spike in the last season we assign a greater weightage to the immediate last period it will give us incorrect results. It has lesser deviation compared to the naïve method.

[pic 3]

Exponential Smoothing:

This uses a smoothing parameter denoted by alpha in addition to the demand for the current period. The value of alpha must lie between 0 to 1. For the FMCG products, the value of alpha is usually on the lower side whereas in case of technological products its on the higher side.

 [pic 4]

The errors are highly reduced compared to naïve method.

Trend Adjusted Exponential Smoothing:

We predict the forecast using the average value and the trend in this case. They are mentioned by alpha and beta respectively. The new average is used to calculate the estimate of trend by taking the difference between averages. The initial requirements of this methods are A0 and T0 which are from the previous data. This is more accurate and takes care in the change in trends and demands.