Part 3- Planning for the SC --> Demand planning Flashcards
Demand forecast
- a complicated and important process
- only been done the last 20-25 years
- the first step in the planning and scheduling process
1) Product or service decisions
2) Capacity planning
Forecasting –> like driving ahead looking in the rearview window
- taking into account what has happened in the near past –> what is the last trends for a particular product or service
- historically forecasts were based on shipments to customers - not a reliable system because instead companies should base their calculations on demands, taking the history of demand as reference and not the quantities shipped
–> often quantities shipped are not the same as quantities ordered by the customers - as technology developed, so did the forecasting methods
The pyramid approach
Bottom up/ top down nr 1 forecast –> uses several statistical methods + other sources of information at various levels of detail
first we get an aggregate number and then the company desegregates to obtain the demands for the different components of the pyramid
The different levels of the pyramid:
The different levels of the pyramid:
1) The product group- the most aggregated level (example softdrinks, a group of products)
2) Product family (according to sales history)- example: Cola drinks (Cola/Fanta)
3) Product/ SKU (Stock keeping unit level) (According to sales history of cola products)- Every SKU has a number and defines a different product from the others- Example: Coke in a can and coke in a bottle has different SKU numbers, it is a different product
4) Customer (According to sales history of cola products in a can)- 45% Merrcadona, 30% Carrefour, 25% Dia
Forecasting realities
1) All forecasts are wrong- it is impossible to have a 100% accurate forecast- companies will try to have them as accurate as possible
2) The more granular the forecast, the less accurate it is- the more specific the more difficult to calculate accurately, then the less accurate it will be/more difficult to get accurate results
3) It is easier to forecast next month more accurately than next year
4) More accurate forecast using demand history (!!!) than sales history- important!!
5) Forecast is a blend of art and science
Types of forecasts
- marketing
- sales
- SC
- finance
Demand drivers
Internal- sales, promotions, discounts
Example: Happy hour in a pub
External- any important changes in the environment- crisis, inflation, war, stock exchange crash
Qualitative models
When a situation is vague and there is little data available (new product). Rely heavily upon intuition and experience.
- Knowledge and intuition of products- Important with the experience, knowledge, the feeling of the manager carrying out the forecast
- Market surveys- improve and modify the products with market surveys to get feedback from potential customers
- Jury of executive opinion- feedback ++
- Delphi method- questionnaires are given to an expert panel anonymously- try to get a compromise and find a correct response through consensus
Quantitative models
When the situation is fairly stable and historical data exists. It is used by current technology products and software. It involves a variety of mathematical techniques.
Time series models: moving averages, weighted moving averages etc
Associate models, causal: Linear and multiple regression
Product lifecycles and forecasting:
Introduction (3D Television) : Qualitative methods
Growth (Blu-Ray players): Qualitative + quantitative methods
Maturity (DVD players): Quantitative methods
Decline (VCR Records) : Qualitative methods
Time series components
Time series can contain some or all of the following concepts:
- trend
- cyclical
- seasonal
- random
Time series models
1) Naive approach- the demand in march has been 10. Then we can say the demand will be 10 in April as well.
2) Moving average- we are in April and we take the demand in january-march as a reference. Take the total demand divided by three and use the average of the three as the demand for april.
3) Weighted moving average- we take jan-march as a reference to calculate the demand in April, but we give different weights to different demands. For example the demand in January is more important because the demand is something cyclical and is more or less repeated every quarter. 50% weight to January, 30% to Feb and 20% to March.
4) Exponential smoothing- the same calculation as the one before but the weights decline with the most recent observation the most weight as it is the most recent observation and therefore most accurate.
Seasonality
Products/services with an important seasonal component.
To calculate:
1) calculate the overall average
2) average each periods historical data
3) divide each periods average by the overall average
4) apply the period index to the existing time series or linear regression forecast
Demand forecasting technology
Done using powerful software applications
