Re: Forecasting Revenue

From: Tim Lee <TLee_at_webcmo.com>
Date: Fri, 22 Jan 1999 18:27:08 -0600 (CST)

MELLIS_at_MAILHOST.ONRAMP.NET WROTE:
>There are so many ways to forecast sales for an e-commerce
>site. I am running some metrics that include the following
>components:
>
>1.) The cost per customer/sale for traditional media,
>affiliate programs and public relations (I already have a
>good idea of what my cost/sale should be for online
>marketing).
>
>2.) The cost per visitor for traditional media, affiliate
>programs (again, I have a pretty good idea of what it should
>be for online marketing).
>
>3.) How repeat business will build in yr. 1, 2, 3 & 4.
>
>Does anyone have any input on what kinds of numbers I can
>use for these components? Does anyone have any research on
>where I can get projections that I can use in my estimates?
>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


Any kind of marketing research (including forecasting) shall
have a purpose. Without knowing your purpose, it is hard to
answer your questions.

On the Internet, there are many forecasts and their only
purpose is to win the favor of news media and get a good
publicity. Under such a circumstance, you don't need to care
about the accuracy of your forecasting. The only rule in
such forecasting is to create the numbers that make good
news.

It is easy to identify such forecasts: there is no error
report about the forecasts. This simple because the
forecasters don't know how to calculate the errors or it is
impossible to calculate the errors or the error is too large
to disclose.

By the way, have we ever seen any error report associated
with any online forecast?

But if you need such forecasted revenue information in your
marketing decisions, you need to know how good your forecast
is. Otherwise, your decisions are quite unreliable.

There are several ways to forecast your revenue:

The first method is to simply use the past revenue
information to make the forecast. The statistical technique
used here is a simple regression. You can get a good report
about your forecasting error by using regression. However,
you need to be cautious: with few observations, your
forecast is always unreliable. (Please see "Let's forecast!"
at http://www.WebCMO.com/forum/fortop4.htm. It demonstrates
the whole process of forecasting and error calculation.)

The second method is to construct a revenue model to
simulate the decision results: if we spend this amount of
the money to promote the site, what will the revenue be
like? I guess Michelle's purpose is this. If it is true, the
purpose of forecasting is not to know the revenue but to
know what is the best investment level. Under this
circumstance, we need to know the cost per sale (or cost per
visitor and the conversion rate) and the retention rate.

Suppose the retention rate is r, cost per sale is cs and the
price is p, your current customer base is m, your annual
promotion budget is b1, b2, b3, b4, and b5 for the year of
1, 2 , 3, 4, and 5 respectively,

Your next year's revenue is: (m*r+b1/cs)*p
Your third year's revenue is ((m*r+b1*cs)*r+b2/cs)*p
.......

As you can see, there are a lot of errors involved in the
parameter (such as r, cs) estimations and there are a lot of
other factors influencing your sales and revenues. So you
need to be very cautious about your forecasting results. It
is interesting to see how your forecasts match your real
revenue five years later.

Talking about forecasting, there is a new product diffusion
model developed by marketing professor Frank Bass about 30
years ago. So this model is also called the Bass model. (The
article about this Bass model is one of a few excellent
marketing research articles I have read: an easy and tidy
model but very useful.) It is used to forecast new product
sales with a few data observations. I believe this model
should be the best for online population forecasts. But I
did not see any marketing research firms using this model in
their online population forecasts.

Best,

Tim Lee
Editor -- Journal of Web Marketing Research
http://www.WebCMO.com
A site dedicated to web marketing research.
WebCMO CyberAnalyzer--understand and target your customers online.


========================================================================
------------------------------------------------------------------------
       This week's Online Advertising Discussion List sponsor:
                                 eGroups.com

 eGroups.com hosts e-mail discussion groups with over 2 million users.
 And you can precisely target them with e-mail advertising that really
 works. To reach opt-in, interested users, contact sales_at_egroups.com

------------------------------------------------------------------------
========================================================================
Online Advertising Discussion List To Unsubscribe send UNSUBSCRIBE
http://www.o-a.com/ to online-ads-request@o-a.com



Received on Fri Jan 22 1999 - 21:47:36 CST