Scribd
Upload
88 views4 pages

Tracking Signal and Method To Control Positive Tracking Signal Biasness

Forecasting is an important tool for predicting future activities and operations. There are various methods to monitor forecast errors over time to improve accuracy, including tracking signals. Tracking signals record predictions and actual figures to identify unexpected deviations. A positive tracking signal means demand is higher than expected, while a negative signal means demand is lower than expected. Consistently positive signals indicate bias in the forecasting method, suggesting the forecasts are under-estimating actual demand. Methods to improve forecasts include using a consistent forecasting model, rolling forecasts, incorporating trends to account for seasonality and variability over time, and better communication across departments.

Uploaded by

Yashashvi Rastogi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
88 views4 pages

Tracking Signal and Method To Control Positive Tracking Signal Biasness

Forecasting is an important tool for predicting future activities and operations. There are various methods to monitor forecast errors over time to improve accuracy, including tracking signals. Tracking signals record predictions and actual figures to identify unexpected deviations. A positive tracking signal means demand is higher than expected, while a negative signal means demand is lower than expected. Consistently positive signals indicate bias in the forecasting method, suggesting the forecasts are under-estimating actual demand. Methods to improve forecasts include using a consistent forecasting model, rolling forecasts, incorporating trends to account for seasonality and variability over time, and better communication across departments.

Uploaded by

Yashashvi Rastogi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

Answer 3

Forecasting is an important tool in operations management. It reflects the art and science of
predicting future activities. The prediction is made in mathematical steps, and if the
probability of deviation is very low, it is considered appropriate. Forecasts are just prediction
which might also go wrong. Forecasts always deviate from the real outcomes.  There are
several ways to monitor forecast errors over multiple time periods to control the forecast.

Tracking Single-

One of these methods to monitor the forecast is called tracking signal. It records the
predictions obtained by comparison with actual figures and warns of unexpected deviations
from the predictions. Forecasts may be related to inventory, sales, or things related to the
company’s future needs. This is a simple indication of prediction bias in the model of
forecast.

The tracking signal is an indicator in which there is prediction distortion or biasness in the
prediction model. It is usually used when questioning the validity of the prediction model.

The tracking signal recognizes all predictions derived from the original comparison and
issues a warning signal when an unexpected deviation occurs in the prediction. This is only
an indication that there is prediction distortion in the forecast model. The tracking signal can
be positive or negative.

A positive tracking index indicates that demand is higher than expected, while a negative
tracking index indicates that demand is lower than expected.

The tracking signal is computed using the following formula-

Here,

MAD denotes the mean absolute deviation which is the difference between the actual demand
and the demand that was forecasted
The cumulative error value may be positive or negative, and the value of the tracking signal
may also be negative or positive. It is calculated that if the value of the tracking signal
exceeds 3.75, it is regarded as a sub-prediction. On the other hand, if the value of the tracking
signal is less than -3.75, it is considered to be over-predicted.

When the actual cumulative demand is different from the forecast, there will be deviations.
The tracking signal can be used to determine the prediction quality. Several procedures were
used. However, one of the simplest methods is to make a total comparison of prediction
errors with absolute deviation.

In the given question when the Tracking signal value is constantly positive that means the
denominator is zero or close to zero always, which is MAD. Also, the positive signal value
depicts that the actual demand is higher than the forecasted demand. By this we can
predict there is some biasness in the procedure and in values.

Also the nature of BIAS is under-forecasted which mean the forecast value is less than the
actual value. There is a negative bias in the forecasting we can estimate this from the above
statement. The data received is consistently too low, which means the forecasted value is
too low as compared to the actual demand.

Methods to improve the Forecast-

1. Using a consistent model-


There is no effective model of forecast that should work effectively in every
organization. Thus, the organization should select one effective and consistent model
for the forecasting process. It is not necessary that an organization use multiple
forecasting models, one effective model can fulfil every requirement. 
And when a single consistent model is used for forecasting of data then the confusion
and variation in result will be less.

2. Implement rolling forecasts-


An organization can update the rolling forecasts based on the present scenario and not
on the result that the management has predicted months ago. With this process, the
forecasting can be done for the next few months and not of the whole year. Rolling
forecasts will help the organization to better align its budget.
And if there is some kind of discrepancy or biasness occurs then it will get removed
earliest.

3. Using Smoothing Constant with Excel solver-


In earlier studies, prediction error measurement was used least with Excel solvers,
using Smoothing constants and introducing values as variables. First use Excel
Solver to complete the optimization, such as Rasmussen (2004).
Since the non-demand period is more than the demand, Solver focuses on minimizing
the error in the non-demand period. This occurs mainly when the fewest errors are
based on MAD. If the demand is low, the zero difference is small, so the forecast is
close to zero or the minimum prediction error is close to zero during most prediction
periods. This leads to a biased prediction, which is much lower than the actual
demand for 5-10 items tested, which can be compared with the results of Taunter and
Duncan (2009).

4. Better communication technique-


The forecasting technique impacts all the aspects of the organization's business, thus,
an organization wants to keep all the lines of communication open with all the
departments throughout the entire process to avoid any kind of hassle. 

5. Additional Measure of Time Series or Adding a Trend Component-


Coefficient of variance (CV) is a general measure to assess whether demand is
difficult to predict. CV is the quotient between the standard deviation and the mean,
and is therefore dimensionless. The higher the number, the more difficult it is to
predict the time series. The resume may require more information to determine
whether the time series is unpredictable. Two different time series with similar
processes may be very different from the vitreous formation method. One reason is
that the resume does not consider the order of requests. Models that randomly
combine high or low demand may have the same resume as models that reflect
seasons and/or trends.
Let us consider a gamma distribution with 40 observations and demonstrate the
limitation with graph showing its projection. Where CV=0.59 and MACs=0.59

Source: Licentiate Thesis (Evolution of Forecasting techniques and errors)


If we project a trend component in this so that the time series should have both season
variability and trend in it. The second time series is as follows where CV=0.56 and
MAC (mean absolute change) changed to 0.15. The trend is injected between the gap
of 30 and 40 periods.

Source: Licentiate Thesis (Evolution of Forecasting techniques and errors)

From the above data we can get a conclusion that when a trend component/line is
been added in the forecasted data, the mean absolute change get decreased and results
in less biased. Where a trend component is a component of time series that represents
variation of low frequency in a time series, and the medium and high fluctuations
have been filtered out.
These are the methods if used consistently throughout the forecasting the result will
be unbiased and more relevant.

You might also like