Re: Affiliate Junk - Do the Math

From: Joel Gehman <joel_at_affiliatehandbook.com>
Date: Tue 25 Sep 2001 13:13:31 -0500

To Wayne Browning, Gail Leino, et al:

Trying to calculate the success or failure of an
initiative - in this case the apparent health of the
entire affiliate marketing industry - on the basis of
the earnings enjoyed by an "average" participant is
doomed to almost certain failure.

Here are a few thoughts.

1) General Problems With Averages
As a rule of thumb 70% of all observations within a
given population will fall "below average" - it's a
function of big numbers. The exception is when there is
an artificial ceiling on the largest observations (e.g.,
mortality, which in modern times rarely strays much
higher than 100 years old).

So even if we did learn that say 100 affiliates earned
an AVERAGE of $2000, we still know basically nothing
about the distribution of those earnings, or its
implications for the success or failure of affiliate
marketing in general.

2) Sampling Errors
Glenn pointed out that his data was for A) one month
in time; B) comprised of 20 merchants; C) represents
less than 2% of Commission Junction's merchants.

Questions: Is the month in question representative of
the entire calendar? Are the 20 merchants representative
of the wide range of industries operating affiliate
programs? Is CJ representative of other providers?

It could be that Glenn's data is completely representative
or not at all. Since there are about 5,000 affiliate
programs today, on the face it seems unlikely that this
set of 20 merchants is representative.

3) Numbers Lie
I recently interviewed the affiliate manager for the
Crucial affiliate program
(http://www.cashpile.com/critic082301.cfm). He indicated
that Crucial moved its program from vendor A to vendor B.
In the process, it went from having 20,000 affiliates at
vendor A, to only 2,000 affiliates at vendor B. However,
total sales volume stayed flat the month after the
move.

Did Crucial's "average" increase 10-fold? Well, sort of,
but what really happened was a bunch of dead wood never
took the time to move from vendor A to vendor B. However,
if we looked at the earnings of individual affiliates,
we would be unlikely to find material differences. So
the fact that the "average" increased is completely
immaterial to the health of this particular program.

Moreover, I would argue that "average" earnings are
equally useless when it comes to measuring the health
of the overall affiliate marketplace. From my first-hand
experience with several dozen affiliate programs over
the last several years I can say that in most cases
affiliate programs work. Of course in some situations,
affiliate programs don't work, but then it is generally
the fault of the operator (e.g., the merchant and/or
affiliate), not the method (e.g., the network or concept).

Joel Gehman
Publisher, AffiliateHandbook.com
ph: 610-792-9597



Received on Tue Sep 25 2001 - 13:13:31 CDT