I currently forecast # of orders for for a catalog based on it's life of 80
days. The forecast is broken out by expected orders per day. I also track
daily actual orders as they come in and compare them to my original forecast.
After 20 days, I would like to re-forecast the likely future orders based
upon my original forecast and the actual orders to date. Any suggestion?
Something from the Analysis Toolpak?

Thanks,
Tom

Pehaps I'm not understanding exactly what you are asking, but what does
your original forecast have to do with future sales mid stream. Said
another way, what is the correlation between any forecast and actual
sales? I would profer the answer is simple, none. I've had a bit too
much coffee today so I will give this a stab. Keep in mind this is
more of a flow of consciousness than a "how to" on sales forecasting.

If you have done a similar catalog in the past and you have daily sales
data, I would assume that there is a definate life cyle to the sales.
Meaning, there would be strong sales early and the closer you get to
day 80 and the further you get from day 1 the sales per day would tend
to drop (on a daily average). If this is true or not, it doesn't
really matter, as long as you know how the age of the catalog
correlates to the number of orders you get per day.

What I would do is look at historical catalog results at Day 20
(cummulative orders) and caculate out a correlation of the subsequent
days. Meaning day 21 is x% of the total of the first 20 days of sales
data, then 22 is y% of the 20 day total, etc. At this point you would
have a working model that you can use against the 20 data set that you
would have in a new campaign and would provide a more realistic
predictor of the sales you can expect in the remaining 60 days.

Now I must say here that there are limitations to this. Primarily you
need to be doing similar campaigns in product, catalog circulation, and
market segment. Secondly, I would assume that there is some
seasonality in catalog marketing, if you can grab a similar campaign
from last year at this time it may be more accurate as a predictor.

Just some thoughts, hope it helps.

PZan


--
PokerZan
------------------------------------------------------------------------
PokerZan's Profile: http://www.excelforum.com/member.php...o&userid=23480
View this thread: http://www.excelforum.com/showthread...hreadid=391319