2. Rather than searching for the perfect forecast, it is far more important to establish the practice of continual review of forecast and to learn to live with inaccurate forecast
3. When forecasting, a good strategy is to use 2 or 3 methods and look a them for the commonsense view.
2. Quantitative – rely on mathematical methods, used when situation is “stable” and historical data exists (existing products, current technology)
2. Independent demand – demand that cannot be directly derived from the demand for other products. Firm can:
a) Take an active role to influence demand – apply pressure on your sales force
b) Take a passive role to influence demand – if a firm is running on a full capacity, it may not want to do anything about demand. Other reasons are competitive, legal, environmental, ethical, and moral.
2. Time series analysis (focus of this chapter) – based on the idea that data relating to past demand can be used to predict future demand
3. Casual relationships – linear regression techniques, assumes that demand is related to some underlying factor or factors in the environment.
4. Simulations – allows the forecaster to run through a range of assumptions about the condition of the forecast.
2. Trend lines – usually the starting point in developing a forecast. Adjusted for seasonal effects, cyclical elements and any other expected events that may influence the final forecast. A linear trend is a straight continuous relationship.
3. Seasonal elements
4. Cyclical elements – more difficult to determine, cyclical influence comes from political elections, war, economic conditions, or sociological pressures.
5. Random variation – caused by chance events. If we can’t identify the cause of the reminder portion of demand, it is assumed to be purely random chance.
6. Auto-correlation – the value expected at any point is highly correlated with its own past values. When demand is random, it may vary widely form one week to another. Where high auto-correlation exists, demand is not expected to change very much from one week to the next.
2. S-Curve – typical of product growth and maturity cycle
3. Asymptotic – starts with the highest demand growth at the beginning then tappers off.
4. Exponential – explosive growth. Sales will continue to increase – assumption that may not be safe to make.
1. Short term – under 3 months – tactical decisions such as replenishing inventory or scheduling EEs in the near term
2. Medium term – 3 M-2Y – capturing seasonal effects such as customer’s respond to a new product
3. Long term – more than 2 years. To identify major turning points and detect general trends.
2. Data availability
3. Accuracy required
4. Size of forecasting budget
5. Availability of qualified personnel.
Y – dependent variable
a – Y intercept
b – slope
X – independent variable
It is used for long-term forecasting of major occurrences and aggregate planning. It is used for both, time series forecasting and casual relationship forecasting.
Decomposition of time series – identifying and separating the time series data into these components.
Trend – easy to identify
Cycle, Autocorrelation & Random Composition – hard to identify.
2. Multiplicative Seasonal variations (seasonal variations is a % of demand) – Forecast = Trend X Seasonal factor
2. Deseasonalize the original data
3. Develop a least sq. regression line for the deseasonalized datat
4. Project the regression time through the period of the forecast
5. Create the final forecast by adjusting the regression line by the seasonal factors.
1. usual errors similar to the standard deviation of any set of data
2. errors because the line is wrong
if you want to forecast June with a 5-month moving average, we can take the average of Jan, Feb, March, Apr, and May.
The longer the moving average period, the more the random elements are smoothed, but lags the trend. Main disadvantage is that all individual elements must be carried as data because a new forecast period involves adding new and dropping the earliest data. In a forecasting model using simple moving average the shorter the time span used for calculating the moving average, the closer the average follows volatile trends
A company wants to forecast demand using the weighted moving average.
Most recent weighting = alpha(1-alpha) na 0
Data one time period older = alpha(1-alpha) na 1
Data two time period older = alpha(1-alpha) na 2
2. Formulating an exp. model is relatively easy
3. The user can understand how the model works.
4. Little computation is required to use the model
5. Computer storage requirements are small because of the limited use of historical data.
6. Tests for accuracy as to how well the model is performing are easy to compute.
– Smoothing constant alpha ()
– Trend smoothing constant delta (δ)
Delta reduces the impact of the error that occurs between the actual and forecast. If both alpha and delta are not included, the trend overreacts to errors.
1. 2 or more predetermined values of Alpha – depending on the degree of error, different values of Alpha are used. If the error is large, Alpha is 0.8, if error is small, Alpha is 0.2
2. Computed values of Alpha – exponentially smoothed actual error divided by the exponentially smothered absolute error.
2. Random Errors – those that cannot be explained by the forecast model being used.
2. Mean Square Error (or variance) – standard error is a square root of a function. Average of Square error.
3. Mean Absolute Deviation – the average forecast error using absolute value of the error of each past forecast. Average absolute error. The ideal MAD is zero which would mean there is no forecasting error. The larger the MAD, the less the accurate the resulting model.
4. Mean Absolute % Error – Average absolute % Error
– is the # of mean absolute deviations that the forecast value is above or below the actual occurrence.
– +- 5 limits are acceptable
– Uses independent variable other than time to predict future demand.
– Any independent variable must be a leading indicator.
(new products or regions)
1. Market Research – looking for a new products and ideas, likes and dislikes about existing products. Primarily SURVEYS & INTERVIEWS
2. Panel Consensus – the idea that 2 heads are better than one. Panel of people from a variety of positions can develop a more reliable forecast than a narrower group. Problem is that lower EE levels are intimidated by higher levels of management. Executive judgement is used (higher level of management is involved).
3. Historical Analogy – a firm that already produces toasters and wants to produce coffee pots could use the toaster history as a likely growth model.
4. Delphi Method – very dependent on selection of the right individuals who will judgmentally be used to actually generate the forecast. Everyone has the same weight (more fair). Satisfactory results are usually achieved in 3 rounds.
2. Joint business planning
3. Development of demand forecasts
4. Sharing forecasts
5. Inventory replenishment
Largest hurtle is lack of trust over complete information sharing btw supply chain partners. Front-end partnership agreements nondisclosure agreements, and limited information access may help to overcome these fears.