Protecting forcast for businesses

[semidevil]

So I work at a retail store, and the way the company works, they determine how many hours are given to each store by the amount of sales that is predicted for this week. The higher number of sales, the higher number of hours = more workers.

the way that the company predict how much we will make this month is this: so let's say the company wants to forcast how much money they will make in the month of August 2005. They take the sales of August of 2002 + Sales of August 2003 and find the average. that average, is how much we are predicted to make in 2005

so (n-2 + n-1)/2 is the prediction for month n..

Being a math major, I was thinking, is this mathematically sound way to predict forcast?

Im sure there are statisticaly better way to predict forcast right? what would be a more accurate way to do it? Since I have access to old records(and have taken a number of statistics and math classes), I was wondering if it is possible for me to give a better prediction.

thanx guys

 

[CRGreathouse]

I would try to find an expected seasonal variation (what % of a year's sales can be expected in a typical January, for example) and then look at the smoothed graph of the sales by month. Perhaps a weighted linear least-square regression would work here -- maybe weight = (30 - number of months ago)?

Really, I'd have to see the data to decide what kind of model I'd use, and I'm sure you can't post that. Still, I agree that a better model should be used here.