Forecasting

[Brisbane Rob]

We budget to manufacture about 10 000 units per month and the dail
production is recorded in a column along side the applicable date. Th
materials we use vary in quality which means production can sometime
run above or below budget for days at a time. What would be the bes
method to forecast production on an on-going basis?

We work Monday to Friday but occassionally on Saturdays.

Any ideas?

Thank

 

[Guest]

The best forecast is based upon specific manufacturing data such as
inventory, quality rejection rates, etc. Another reasonable approach is to
use a mirror forecast:

Keep track of two separate running averages. The short-term average over
the previous 10 days and a long-term average over the preceding quarter.

Forecast the next 10-20 days using the short-term average and the balance
forward using the long-term average.

TO validate this approach, keep track of predicted production vs actual
production

 

[Brisbane Rob]

Thanks, I'll give it a bash. Much appreciate

 

[Conrad Carlberg]

Moving averages of the sort that Gary's Student recommends are certainly one
possibility, but hardly the only one and depending on your baseline data
that approach could be entirely wrong. (BTW, you can find a moving averages
tool in the Analysis ToolPak [sic] which comes with Excel.)

Two other approaches are exponential smoothing and regression, both of which
are also included in the Analysis ToolPak. There are variations such as Holt
and Holt-Winters, but they're not offered directly in Excel -- you have to
do it yourself.

As Gary's Student recommends, you should track your forecasts against your
actuals, and perhaps choose your method on that basis. But the best method
is to use ARIMA analysis (Autoregressive Integrated Moving Averages, or
Box-Jenkins analysis), which is an objective method of determining whether
to use moving averages, smoothing, regression, or a combination of those
ingredients. It also informs you as to seasonality in your baseline.

And you should also investigate whether your baseline involves a trend,
whether up or down. If it does, you want to start by differencing the data
so that it becomes stationary.

--
C^2
Conrad Carlberg

Excel Sales Forecasting for Dummies, Wiley, 2005


"Brisbane Rob" <[email protected]>
wrote in message
news:[email protected]...