Professional Documents
Culture Documents
Supply Chain Planning Notes
Supply Chain Planning Notes
v) It can be represented as
∑ Di
F t= i=1
t−1
c) Moving Average –
i) The moving average (MA) is a forecasting method where the average is taken over a
specific time period, like 10 days, 50 days, 20 weeks or any time period we choose.
This smooths out demand data by creating a constantly updated average price.
t−1
∑ Di
ii) It is represented as i=t− N+ 2
F t=
N
iii) Here, N is the number of periods you're going to average together, which we can
choose. For small values of N, the forecast is more sensitive but also picks up more
noise and therefore is quite uneven. A large N makes is more stable, but also less
reactive (to recent changes or patterns). You have to try different values of N,
measure accuracy for each one of them, and then pick the best one.
iv) Moving average is a tremendously adaptable forecasting method. You can make it
very reactive or you can make it very stable. And that's why a lot of companies that
have reasonably stable demand use it.
d) Exponential Smoothing –
i) Exponential smoothing, similarly to the moving average, is a very versatile method.
By changing one value (alpha), you can make it more reactive or more stable.
ii) It is represented as
iii) Alpha goes from 0 to 1 and we have to vary it to get the most accurate model. If your
accuracy is improving, then you're going in the right direction. If your accuracy's
getting worse, you're going in the wrong direction.
iv) There are a number of extensions to the exponential smoothing formula. You can
take into account trended data, which means you have a consistent increase over
time or decrease over time. Or you can estimate demand that is seasonal.
v) Just Trend – Holt’s Model
Both Trend and seasonality – Winter-Holt’s Model
d) Mean Absolute Percentage Error (MAPE) - Forecast accuracy in the supply chain is
typically measured using this. Statistically MAPE is the average of percentage errors.
D−F
It is defined as
MAPE=∑
| D |
t
e) Mean Squared Error (MSE) – MSE is used because it gives more weightage to larger
errors because it squares them. This is because larger errors are to avoided at all costs
as they can disrupt our supply chain more and make planning more difficult.
( D−F )2
It is defined as MSE= ∑
t
f) When deciding on which forecast is most accurate, we should compare all the three
measures as each of them tell us something different.