Midterm #1 Flashcards
Demand Planning
key building block from which all supply chain planning activities are derived from
Demand Planning Approach
1)review accuracy, understand past data
2)develop statstical forecast
3) add market intelligence
4) agree in consensus mtg
Fundamentals of forecasting
1) forecast wrong
2) simple methods»> complex
3) a correct forecast does not = correct forecast method
4) don’t use data regularly trust it less when forecasting
5) all trends will end
6) forecast is bias
7) tech does not equal better forecasting
level
average rate of sales
trend
the growth or decline of product
seasonality
repeating patterns in demand
events
one-time activties that influence demand
noise
randomness in the infromation
cumulative error
sum of forecast error between each actual demand and forecasted value
average error
cummulative error/n
Mean Absolute Deviation (MAD)
average of absolute deviation between each actual demand and forecasted value
Mean Absolute Deviation (MAD)
average of absolute deviation between each actual demand and forecasted value
MPE
(actual - forecast /actual*100)/n
MAPE (mean aboslute % error)
average of all the absolute % errors
forecasting bias
tendency to over or under forecast consistently over time
time series forecasting
set of data obtained by making observations at equally spaced points in time
Naïve Model
used when no trends or seasonality are present assumes most recent demand is the best estimate of the next value of demand.
-last period model (Ft=Dt-1) *fluctuation not random
- arithmetic mean (average of Demand) *fluctuations that are random
smoothing
for a stationary time series with no trend we need to “smooth out” random fluctuations
moving average
is the average of the (n weeks) in demand before forecasted value so if it says 3-week moving average you are forecasting 4th value but using 3 previous values
Moving average- choice of N
n is chosen large enough to include enough observations to smooth out random fluctuations but small enough not to give weight to irrelevant past info, smaller n will track shifts in the level of a time series faster
large n- past info, smoothing
small n- recent data
weighted moving average
If the most recent periods are weighted too lightly, the forecasted values will assume that actual changes in demand are random.
0.4, 0.3, 0.2, 0.1
ABC system
classifies inventory based on the revenue contribution of an item highest -least value
80/20 rule
20% of skus make 80% of sales (A iteams)
80% of skis make 20% of sales (B and c Iteams)
b iteams monthly, c quarterly reviews
exponential smoothing
weights all past values of demand where the weights decrease geometrically with increasing age of data
Ft+1 = αDt + (1- α)Ft
Choice of α
Select α between 0 and 1
large Choice of α give more weight to recent values; greater sensitivty to variation
0.1 < α < 0.2 or 0.3
*on excel data-data anaysis- damping facor = 1-α
EXPO SMOOTING TREND FACTOR
T(t+1) = β(F (t+1) – Ft) + (1 – β) Tt
β = smoothing constant for trend; 0.0 ≤ β ≤ 1.0
Adjusted Expo
AF(t+1)= F(t+1)+T(t+1)
seasonal factor
period demand/ year average
deseasonalize data
sales/ seasonal index
seasonal index
average of seasonal factors
trend values
T=a+bx
single regression
based on a sample of n data points for two variable X and Y we from a model to predict the dependent variable y as a function of one independent variable x
correlation r
the measure of the strength of the relationship between y and x
-1 ≤ r ≤ 1
coefficeint of determination r^2
measure the proportion of the total variation from the mean explained by the regression model 0 ≤ r^2 ≤ 1
assumption based forecasting
when forecasting new product with no past data
