demand forecasting Flashcards
forecasting
consistent, systematic and appropriate forecasting process positively impact performance through decreased operations costs, improved customer service, increased sales, and reductions in inventory. these improvements have positively affected return
what is forecasting?
a method for translating past experience and present events into predictions of the future
observed measurement = systematic part + random part
- forecasting tries to isolate the systematic part
- the random part determines the forecast accuracy
why do we forecast
-to make better decisions
forecasting is important because:
- it is a starting point for business planning
- all business decisions will follow the result of forecasting
- “bad “ forecast can lead to a significant increase in cost
Forecasting is vital to every functional area:
- finance
- marketing
- HR
- production and operations
forecasting error
forecasts are rarely perfect; actual results differ from predicted values
forecasting techniques generally assume that:
the same underlying causal system that existed in the past will continue to exist in the future
Forecasts for ______ items tend to be more accurate than forecast for ______ items
- group of
- individual
_________ forecasts tend to be more _________ accurate than forecasts
- Short-term
- long-term
forecasting methods
- judgmental methods
- time series analysis
- causal relationship
- simulation
judgmental methods:
- sales force estimate
- executive opinion
- market research
- historical analogies
- Delphi method
Delphi method:
Group of experts respond to questionnaire.
A moderator compiles results and formulates a new questionnaire that is submitted to the group
time series:
a time ordered sequence of observations taken at regular intervals of time
assumptions
(i) record of past demand is available and (ii) past demand is a predictor of future demand
the following 5 patterns could be identified in a time series:
- trend
- seasonality
- cycles
- irregular variations
- random variations
trend
long term movement in data (up or down)
seasonality
short-term fairly regular variations in data generally related to factors such as the calendar or time of the day
cycles
wavelike variations of more than one year’s duration. These are often related to a variety of economic, political and agriculture conditions
irregular variations
-caused by unusual circumstances such as severe weather
random variations
-caused by chance that will remain after all other behaviours have been accounted for
fundamental difference between cycles and seasonality is the …
duration of repeating patterns
linear regression
Is a way to model the relationship between two variables. …
The equation has the form Y=a+bt,
- where Y is the dependent variable (that’s the variable that goes on the Y axis),
- t is the independent variable (i.e. it is plotted on the X axis),
- b is the slope of the line and a is the y-intercept
Naïve Method
the Naïve forecasts are simply set to be the value of the last observation
Naïve Method -9 facts
- Simple to use
- This method works remarkably well for many time series.
- Very low cost
- Data analysis is nonexistent
- Easily understandable
- Can serve as a standard for comparison
- Low accuracy (not suitable when the random variation is high)
- Short-term
- Providing a starting point for other forecasting method
Uses for naive forecasts
- with a stable time series data
- with seasonal variations
- for data with trend
with a stable time series data
the last point becomes the forecast for the next period. if the actual demand for a product last week was 20 cases, the forecast for this week is 20 cases
formula + example
With a stable time series data
with seasonal variations
the forecast for this “season” is equal to the value of the series last “season”. for example, the forecast for demand for energy this summer is equal to the demand for energy last summer
formula + example
With seasonal variations
for data with trend
the forecast is equal to the last value of the series plus the difference between last two values of the series. for example, suppose the last two values were 50 and 53
formula + example
fot data with trend
techniques for averaging
- moving average
- weighted moving average
- exponential smoothing
moving average
-averages a number of the recent actual data values which updated as new values become available
weighted moving average
assigns large weight to the most recent values in a time series
exponential smoothing
is sophisticated weighted averaging where a new forecast is based on the previous forecast plus a percentage of the difference between that forecast and the previous actual value
simple (moving) average
- relies on the recent past data
- smooth out random variations
- useful for stable demand
simple (moving) average formula
N= total number of periods used in calcuation
- sensitivity to random variation depends on N
- N is subjective
Compute a three period-moving average forecast for period 6, given for the demand below.
the number of periods (N) to include in simple moving average
- Arbitrary
- The fewer, the more sensitive (responsive) to most recent data.
- If responsiveness is important, using relative few data. Keep in mind, it also cause the forecast to be responsive even to random variations.
advantage of simple moving average
- easy to compute
- easy to understand
disadvantage of simple moving average
- values in the average are weighted equally
- can not pick up trends very well
- may require extensive records of past data
weighted moving average
- weights represent the amount of influences of past demand on forecast
- weights reflect fluctuations in the demand data
weighted moving average formula
- Wt-1 = weight applied to period t-1’s demand
- weights are non-negative and sum to one
- most common: most current demand has larger weight
- weights and N are subjective
weighted moving average problem
- calculate a weighted average forecast using a weight of 0.40 for the most recent period, 0.30 for the next most recent, 0.20 for the next and 0.10 for the next
- If the actual demand for period 6 is 39, forecast the demand for the period 7 using the same weights as in part 1.
- ) F6 = 0.40(41)+0.30(40)+0.20(43)+0.10(40) = 41.0
- ) F7 = 0.40(39)+0.30(41)+0.20(40)+0.10(43) = 40.2
exponential smoothing - facts
- is is a special case of the weighted averages method in which we select only one weight-the weight for the most recent observation
- the weights for the other data values are automatically computed and get smaller as the observations move farther into the past
- requires minimal amount of data
exponential smoothing formula
exponential smoothing example
exponential smoothing example 2
- use alpha = 0.1 find the forecast for period 6
- use alpha = 0.4 find the forecast for period 8
picking a smoothing constant – criteria
- selecting a smoothing constant is basically a matter of judgement of trail and error . the goal is to select alpha that balances the benefits of random variations
- commonly used values of alpha range from 0.01 to 0.5
- if the real demand is stable we would like small alpha to lessen the effects of short term or random changes
- if the real demand is rapidly increasing or decreasing we would likea large alpha to try to keep the changes
the most used of all forecasting techniques
exponential smoothing
exponentially models are surprisingly _______
accurate
___________ exponential model is relatively ________.
- formulating
- easy
with exponential smoothing, the user can ___________ how the model ________
- understand
- works
with exponential models, little ___________ is required by to use the model.
computation
with exponential smoothing, __________________ requirements are small because of the limited use of ______________ data.
- computer storage
- historical
with exponential smoothing, _______ of _______ as to how well the model is performing are _____ to compute
- tests of accuracy
- easy
in exponential smoothing method, only three peices of ata are needed to forecast the future:
- the most recent forecast
- the actual demand that accrued for that forecast period
- smoothing constant alpha
linear trend equation - Summary
- fits a trend line to a series of historical data
- if we want to use a precise statistical method we can use least square method to find the equation of the line (called the least squares line)
linear trend equations
the coefficients of the line, a and b can be calculated from data using these two equations
n= total number of periods
y=value of the demand time series
linear trend example
the manager of Hi-tech company wants to develop a forecat model for annual sales of laptops. the following data have been collected for five different periods. Develop a linear trend equation for these data, then forecast sale for week 10.
forecasts are rarely perfect :
- internal factors or external factors
- more product choices, faster introduction of new products
- lack of data, lack of expertise, time or money constraint
forecasting objective: to minimize forecasting error
forecasting error = (actual demand) - (forecast)
to minimize error :
- select the best possible model
- select the best possible value for a parameter (N in Moving Average, alpha in Exponential Smoothing)
Minimize forecast error measures:
- MAD
- MSE
- MAPE
MAD
Mean Absolute Deviation
- advantage- based on absolute values, the errors of opposite signs do not cancel each other out when they are added
- measures general variation
- smaller the better
MSE
Mean Squared Error
- Advantage- Due to the squaring of the error team, large errors tend to be magnified
- reveals large errors
- smaller the better
MAPE
Mean Absolute Percent Error
- advantage - weights error relative to actual values
- smaller the better
Monitoring forecast errors
tracking signal formula
When tracking signal values exceeds 4 in absolute value, the analyst will study the model and make changes. the model may be in adequate due to :
- the omission of an important variable
- a change or shift in the variable that the model cannot deal with
As long as the tracking signal is between -4 and 4, assume the model is working correctly
