demand forecasting

Question 1

what is forecasting ?

Answer 1

• anticipation of demand based on the past

- look for correlation in variables to make forecasts

Question 2

the two types of demand :

Answer 2

INDEPENDENT DEMAND = non controllable demand / driven by the market
based on the sales of the final product
the black line of the graph = what we want to forecast

DEPENDENT DEMAND = controllable (the component are designed by the company)
manageable
depend by the independent demand

Question 3

different level of forecasting (classifing them in ling term, medium, and short)

Answer 3

the farther is the less detailed will be :

     - LONG TERM:  independent demand   -    general 
                               BUSINESS FORECASTING :  esteem of the number of item sell (long period)  
                                                                                 general forecast ( no model, feture ...) 

 - MEDIUM TERM :  aggregate forecast
                              AGGREGATE :  aggregation of different family of product similar to each other 
                                                        to optimize the production 

 - SHORT TERM : operative productiond organization = DEPENDENDT FORECAST
                             ITEM : Detailed forecasting of all the items nedded in the production
                             RAW : all the rae matirial neded in for the production

Question 3

The principles of forecasting

Answer 3

• forecast are wrong = HISTORY IS NOT A PERFECT PREDICTOR OF THE FUTURE
• better for family of product
• better in the short term ( long term there are too many variables)
• better for larger areas
• should include an error estimate

Question 4

The two way to forecast

qualitative and quantitative

Answer 4

• qualitative = relay on opinions of expert, passed experience
• quantitative = analyzing data with different strategies

Question 5

qualitative way to forecast

Answer 5

• MARKET RESEARCH: try to customers to identify their habits –> derive future demand
• HISTORICAL ANALYSIS : analyze past situations to get similarities in the market
• PANEL OF EXPERT: asking multiple experts of different sectors of my company
  - different point of view / suggestions
  - also experts can be wrong
  - difficult to reach a full agreement
• DELFI METHOD: a panel of experts that answer questionary (one or more sets) (anonymus)
  collect the answers, evaluate them, ask confermations, if there are opposite
  points of view ask to do another survey and articulate theyr ideas, different
  scenarios and solutions (very dettailed)

Question 6

trends are affected by

Answer 6

seasonality: is important to identify them and don’t compare 2 different periods
random fluctuations: if they are out of the trend they must be discarded (if i have enough data)
or smoothed
è l’unico modo per annullare questo non c’era del altro ?

Question 7

moving averages

Answer 7

MA is a simple way to look at the past data of demand and predict the future

SIMPLE MA : somma di N dati / N

		- Più N (punti o periodo) è grande più è difficile da muovere
		- Period usually 4-9 
		- Smooth the random variations = cuts picks and valley 
		- Problems :   -  give to much importance on past value 
	                               - Always lags in time 

weighted MA: non fatta ma da più importanza al dato più recente in base a dei pesi assegnati

EXPONENTIAL SMOOTHING MA:
Balance the problems of the MA										
Composed by 2 factors : 
					- Sales in the previus period 
                                        - Forcast of the previus period 

    Alpha = smoothing costant           	 Weight of the lastest data 
		Determines the responsiveness to change in demand 													 
			Alpha = 0    the new forecast is = to the last  													
			Alpha = 1   the new forecast = last sales 												
	        	I need to choose apha in a way it is rappresentig my data 
      			- Too small = doesen't change with the graph 
                	- Too big = too reactive in the change of the graph 
                           I choose it whatching the graph = what better forecast

Question 8

difficlult relation btw variables / casual relationship / new point based on the past

Answer 8

Casual relationship: stat. techniques to establish relations between demand and different variables

• LINEAR REGRESSION = for casual relationship
  A way to identify a specific value based on the past
  The position of a new point in the future = dependent on the position of the past points
  I must have a linear trend (i can linearize exp and log trends (?) )
  Useful to understand complicated relations btw variables

si basa su un equazione