Characteristics of forecast

1. Forecasts are always inaccurate and should thus include both the expected value of the
forecast and a measure of forecast error. Example forecast between 100 to 1900 and 900 to
1100 have average as 1000 but the error is higher on the former side.
2. Long-term forecasts are usually less accurate than short-term forecasts; that is, long term
forecasts have a larger standard deviation of error relative to the mean than short-term
forecasts. Therefore, firms try to keep replenishment lead time as low as possible as it helps
the store manager to forecast for a short-term period.
3. Aggregate forecasts are usually more accurate than disaggregate forecasts, as they tend to
have a smaller standard deviation of error relative to the mean. Example, it is easier to
forecast GDP with less error but not the revenue of a company.
4. In general, the farther up the supply chain a company is (or the farther it is from the
consumer), the greater the distortion of information it receives. (bullwhip effect).

Forecasting take into accounts the following:

• Past demand

• Lead time of product replenishment

• Planned advertising or marketing efforts

• Planned price discounts

• State of the economy

• Actions that competitors have taken

Types of forecasting methods:

1. Qualitative: Depends on human judgement

2. Time series: Using of historic data for forecasting.
3. Casual: Developing a co-relation with environment factors (Economy, interest rates etc.,)
4. Simulation: Here we check the customer behaviour with respect to price changes, promotion
etc., and do forecasting.

Time series Methods

Observed demand (O)= systematic component (S) + random component (R)

Systematic component: The systematic component measures the expected value of demand
and consists of what we will call level, the current de-seasonalized demand; trend, the rate of
growth or decline in demand for the next period; and seasonality, the predictable seasonal
fluctuations in demand.

Random component: The random component is the part of the forecast that deviates from the
systematic part. A company cannot (and should not) forecast the direction of the random
component. All a company can predict is the random component’s size and variability, which
provides a measure of forecast error.

The forecast error measures the difference between the forecast and actual demand.
Approach to forecasting

1. Understand the objective: All the stakeholders should be aware of the reasoning behind
(basis of) a forecast. Whether it is a forecast for seasonal, promotional etc., demand.
2. Integrate forecasting throughout the supply chain: Otherwise there can be a huge
coordination failure for example in case of promotions.
3. identify major factors that influence the demand forecast: Seasonality, presence of
substitutes, delivery lead times etc.,
4. Forecast at the appropriate time: Do not forecast to early, forecast at appropriate time to
ensure the reduction in error.
5. Establish performance and error measures for the forecast: Should measure the forecast
errors to minimize the same in future.

Time series forecasting methods

The equation for calculating the systematic component may take a variety of forms: Systematic
is calculated and random is estimated.

• Multiplicative: Systematic component = level * trend * seasonal factor

• Additive: Systematic component = level + trend + seasonal factor

• Mixed: Systematic component =1level + trend2* seasonal factor

Systematic component is calculated by two methods:

1. Static methods: A static method assumes that the estimates of level, trend, and seasonality
within the systematic component do not vary as new demand is observed. In this case, we
estimate each of these parameters based on historical data and then use the same values
for all future forecasts.

Systematic component = (level + trend) * seasonal factor

St = estimate of seasonal factor for Period t

Dt = actual demand observed in Period t

Ft = forecast of demand for Period t

In a static forecasting method, the forecast in Period t for demand in Period t + l is a product
of the level in Period t + l and the seasonal factor for Period t + l. The level in Period t + l is
the sum of the level in Period 0 (L) and (t + l) times the trend T. The forecast in Period t for
demand in Period t + l is thus given as

Ft+l = [L+ (t + l) T] St+l

We now describe the following two steps required to estimate each of the three
parameters—level, trend, and seasonal factors.

1. De-seasonalize demand and run linear regression to estimate level and trend.

2. Estimate seasonal factors.

 DS= L + Tt

Here, Ds is the de-seasonalized demand, L is the level at period 0 and T is the trend from t=0
to t=t.

Plot this on the graph and get the values of L and T from the regression line and find de-
seasonalized demand for different periods.

Seasonal factors for each month, S = D/Ds

If the seasonality is for three months for 12 months period than cycle of each period is 4.

Therefore, S1=S5=S9 if not then consider the average seasonality S1=S5=S9=S13=…. =

(S1+S5+S9)/3
Therefore, actual demand will be calculated by multiplying the de-seasonalized demand
from regression line and then we can multiply it from seasonality factor.

Adaptive forecasting

In adaptive forecasting, the estimates of level, trend, and seasonality are updated after each
demand observation. The main advantage of adaptive forecasting is that estimates
incorporate all new data that are observed. We begin by defining a few terms:

Lt = estimate of level at the end of Period t

Tt = estimate of trend at the end of Period t

St = estimate of seasonal factor for Period t

Ft = forecast of demand for Period t (made in Period t − 1 or earlier)

Dt = actual demand observed in Period t

Et = Ft – Dt = forecast error in Period t

In adaptive methods, the forecast for Period t + l in Period t uses the estimate of level and
trend in Period t (Lt and Tt respectively) and is given as:

Ft+l = (Lt) + (lTt)*St+l

1. Initialize: Compute initial data for L, S and T as per the previous method.

2. Forecast: Using above method and the value from the first step.

3. Estimate error: Calculate the error between the forecast and actual demand

4. Modify future estimates: By upsizing or downsizing the forecast on the basis of the sign of
the forecast error.

There are different types of adaptive methods:

1. Moving Average: Here trend = 0 and seasonality = 0. Therefore, systematic component

S= Level (L). Lt = (Dt + Dt-1 +…... + Dt-N+1)/N
2. Simple exponential smoothening: Here trend = 0 and seasonality = 0. Therefore,
systematic component S= Level (L). Initial estimate of L0 is taken by moving average.
After observing the demand, Dt+1, for Period t + 1, we revise the estimate of the level as
follows: Lt+1 = α (Dt+1) + (1 – α) -Lt
 3. Trend-corrected exponential smoothing (Holt’s Model): Here only seasonality = 0.
Therefore, systematic component S= Level (L)+ Trend (T). here initial value of L0 and T0 is
taken with the help of linear regression.

In Period t, given estimates of level Lt and trend Tt, the forecast for future periods is
expressed as Ft+1 = Lt + Tt and Ft+n = Lt + nTt.

After observing demand for Period t, we revise the estimates for level and trend as follows:
Lt+1 = αDt+1+(1 – α) (Lt + Tt)

Tt+1 = β (Lt+1 - Lt) + (1 – β) Tt

where (0 <α<1) is a smoothing constant for the level and b (0< β< 1) is a smoothing constant
for the trend.

4. Trend and seasonality corrected exponential smoothing (winter’s model): Here,

Systematic component of demand = (level + trend) * seasonal factor. We obtain the
initial estimates from the static method for L0 , S0, and T0.
On observing demand for Period t + 1, we revise the estimates for level, trend, and
seasonal factors as follows:
Lt+1 = α(Dt+1/St+1) + (1 – α) (Lt + Tt)
Tt+1 = β (Lt+1 - Lt) + (1 – β) Tt
St+p+1 = ϒ(Dt+1/Lt+1) +(1- ϒ) St+1

Measure of forecast error

1. Take measures in case of +ve or -ve error in terms of under-sizing or oversizing. E=F-D
2. Prepare for contingency planes in case of forecast failures. Like having a supplier with a shorter
lead time.

Terminologies in calculating error.

1. One measure of forecast error is the mean squared error (MSE): it is a good idea to use the MSE
to compare forecasting methods if the cost of a large error is much larger than the gains from
very accurate forecasts.
2. Define the absolute deviation in Period t, At, to be the absolute value of the error in Period t;
that is,

3. The mean absolute deviation (MAD) to be the average of the absolute deviation over all
periods, as expressed by

MAD is a better measure of error than MSE if the forecast error does not have a symmetric
distribution. Even when the error distribution is symmetric, MAD is an appropriate choice when
selecting forecasting methods if the cost of a forecast error is proportional to the size of the
error.
4. The MAD can be used to estimate the standard deviation of the random component assuming
that the random component is normally distributed. In this case the standard deviation of the
random component is:

5. The mean absolute percentage error (MAPE) is the average absolute error as a percentage of
demand and is given by:

The MAPE is a good measure of forecast error when the underlying forecast has significant
seasonality and demand varies considerably from one period to the next.

6. Bias: One approach is to use the sum of forecast errors to evaluate the bias, where the following
holds:

7. The tracking signal (TS) is the ratio of the bias and the MAD and is given as
 If the TS at any period is outside the range +-6, this is a signal that the forecast is biased and
is either under forecasting (TS < -6) or over forecasting (TS > +6). This may happen because
the forecasting method is flawed or the underlying demand pattern has shifted.

Setting smoothening constant

In general, it is best to pick smoothing constants that minimize the error term that a manager is
most comfortable with from among MSE, MAD, and MAPE. In the absence of a preference among
error terms, it is best to pick smoothing constants that minimize the MSE.

Forecasting in practice

1. Collaborating in building forecasts

2. Share only the data that truly provide value.
3. Be sure to distinguish between demand and sales.