**Trend-Adjusted Exponential Smoothing **

The exponential smoothing approach discussed above is an appropriate forecasting technique, if the time series exhibits a horizontal pattern (i.e. No trend) with random fluctuations. However, if the time-series exhibits trend, forecasts based on simple exponential smoothing will lag the trend. In such cases, a variation of simple exponential smoothing called the trend-adjusted Exponential smoothing can be used as a forecasting technique. "The trend-adjusted forecast (TAF) has two components:

**A smoothed error****A trend factor**

**TAF**_{t}** = S**_{t-1}** + T**_{t-1}** , where **

S_{t-1}= Previous period smoothed forecast

T_{t-1}= Previous period trend estimate

TAF_{t}= Current period's trend-adjusted forecast

S_{t }= TAF_{t }+ (A_{t}- TAF_{t})

T_{t}= T_{t-1}+ (TAF_{t}- TAF_{t-1}- T_{t-1}),where and are smoothing constants

In order to use this method, one must select values of and (usually through trial and error) and make a starting forecast and an estimate of the trend" (Stevenson, 2005).