Decomposition: No; this is not about reverse engineering or how to make your own organic soil improvement.

When You have time related data and would like to see if there is a trend it can be difficult if there are seasonal variation in the data.

It could be You are investigating if sales of Your product are actually increasing, or marketing is inflating their success in the latest report.

Or it could be a situation where seasonal variation is important is when You try to forecast purchase needs, but You must take high and low season into account to optimize warehouse capacity.

A useful tool whether you are supply chain manager, CFO or chairman of the board guiding the management in the right direction.

A very practical example where the forecast is about periodical costs is the electrical bill, I get every quarter. Winter is of course more expensive than summer, but how much?

Did I save any money since last quarter or last year by getting energy saving appliances and how much can I expect to pay next quarter and next year?

These are the last 10 payments. Each of them represents one quarter of the year. Can You spot any clear trend from this? Probably not, except the first one seems a bit high.

This is what decomposition can do for You.

Without too much technical detail what You see is an analysis that shows a clear tendency to lower cost every quarter when You consider that some quarters are simply more expensive than others – from 2300 above to 2700 below average.

Using the model to forecast future payments told me to expect a bill next quarter of kr. 2,643.37 and I was happy to see an almost perfect match between model and real life when I had to pay kr. 2,660.93

So, for many situations where data are a bit “blurred” from seasonal variation a decomposition analysis can paint a clear picture for You and improve Your predictions.

There will be more interesting stuff later, so stay tuned.

In the meantime, have a look at my other interesting posts about process validation, sample size determination and risk analysis here.

For more information please contact EpsilonPlus here.