Please enable JavaScript.
Coggle requires JavaScript to display documents.
Production Planning and Control - Coggle Diagram
Production Planning and Control
Week 1 - Forecasting: Gelecek için tahminde bulunma
Time Horizons
Medium-range foreast: betw. 3 months to 3 years
Long-range forecast: more than 3 years, şirket üzerine uzun süreçte etkisi daha fazla olan şeyler.
Short- range forecast: Up to 1 year, generally < 3 months, daha quantitative işler.
Types of forecasts
Demand forecasts
Technological forecasts
Economic forecasts
Seven Steps of Forecasting
Where will you use the forecast?
What will be forecasted?
What is the time horizon?
Which forecasting model(s) will be
Data collection
Forecasting
Validation and implementation
Must Know
Forecasts are generally wrong
They have assumptions
Aggregated forecasts are more accurate than individul ones. (Tek bir ürüne odaklanmak iyi sonuç vermez.)
Forecasting Methods
Qualitative Methods - based on personal opinions in desining new products or new technologies. They are subjective.
Jury of executive opinions: A small groups of executives come together to make some insights with using their experiences. Combining of executive opinions with statistical models. If there is a dominant executive, the decision making process can be shaped by this one.
Delphi Method: The experts answer questionare about execution project. The aim of this method is prevent the influence of dominant ones.
Sales force composite: Salespeople forecast the sales in the future. Because they know the customers best But they set the goals for the company lower, because they have to achieve these goals.
Consumer Market Survey: Ask customers what they prefer. It is substantial to select the correct samples.
Quantitative methods - based on mathematical models in objective perspective to cultivate existing products and current technologies.
Time-series model. Assumption: Future demand only depends on past demans. Data are collected in regular time scales.
Naive Approach: The demand in the future will be same as the past demans. Can be a good start. cost effective for some products.
2.Moving Avarages: demand forecast for the future period will be avarage of the n period. data must be stationary(everything is constant) Sum pf the demand as past periods/n
Trend
, persistent(kalıcı), upward or downward pattern
Cyclical
Data shapes like a wave. There are some ups and downs.
Not regular
Seasonal
Data tends to act like every same period of a time scale. Repeats itself after a certain period.
regular
Random
Data acts randomly, it will not be predicted.
Weighted Moving Avarage: We multiply the past demands with their distance with inverse proportion (ters orantı) and divide the sum of the weightnesses. For example: we are trying to find april's demand, we multiply march's demand with 3, feb with 2 and october with 1. after that we take the sum of these datas. After that divide by 6.
Challenges: we are not able to use these data in trends. with moving average. when we increase n, the forecast is more smoothes but less sensitive to changes. we should same period of the data.
Exponential Smoothing: Data weighted averages decrease towards the past. We need to identify a smoothing constant. 0<a<1. Needs to recording the only past sales and data. we can not use this method in trends. New forecast = a
Actual demand of last month + (1-a)
Forecast for last period
It is not work well with trends. If we decrease the a, there is a more smooth forecast, but it is less sensitive to changes. As a increases the weighted of the recent data increases
Associative Model