March 29th, 2006, 02:01 AM
|
|
Affiliate Manager
|
|
Join Date: February 25th, 2005
Location: Ventura, California
Posts: 117
|
|
Slave to the CJ EPC - When a Well-Meaning Statistic Takes on a Life of its Own!
EPC, the statistic an affiliate program lives and dies by at CJ. The relentless pursuit of a sky high EPC can lead to some pretty wacky behavior on the part of program managers, and is actually in direct opposition to the goal of profit maximization for an affiliate program (assuming people would behave rationally, which they don't).
First, a quick refresher on the principle of profit maximization, straight out of Econ 101. Recall the darling of all micro economic courses, the widget maker. The widget maker should keep making additional widgets until the marginal cost of producing the nth widget is equal to the marginal revenue that can be obtained through the sale of the nth widget. As long as Widget.com is making just a teeny bit of profit by making that extra widget, it makes sense for them to do so.
Fine you say, but what does that have to do with the price of guns and butter? Well, nothing, but you can apply the profit maximization principle to affiliate marketing, in the context of EPC, to see what Adam Smith would say about program management and the proper strategy that would lead to profit maximization for an advertiser's affiliate program.
So what is marginal cost and marginal revenue in the context of EPC?
Let's take a hypothetical, established program, TeleBling.com (TB) that has thousands of affiliates and is cruising along generating a satisfactory level of revenues. Affiliate management experts come along and tell TB that in order to grow their program they must attract affiliates, and to do that, they have to increase their EPC, and the best way to do that, they tell them, is to terminate all of those nasty affiliates who have the audacity to refer them traffic that does not convert to sales at the rate needed for their target EPC.1 The affiliate manager then dutifully rushes off and sends the ole "We regret to inform you that..." (CJ's version of the Dear John letter) to hundreds, possibly thousands of unfortunate affiliates! Because, "As you know, EPC plays a critical role with the CJ network...."
So why is this a problem? Well, let's apply the marginal cost vs. marginal revenue concept, i.e. (PROFIT maximization, not EPC maximization).
What is the actual additional cost of keeping an affiliate in your program?
Well you say, you have to look at them in the reports and they confuse your affiliate manager and make him/her grumpy and less efficient! OK, so group your affiliates into "Those I care about" and "Those I don't" and then instead of terminating them, hide the "Those I don't" group from your affiliate manager. Problem solved.
But, you say, there is a real marginal cost involved when sending email newsletters through the CJ system! True, and that is a real marginal cost! Please refer to my solution to problem 1; Don't email the "Those I don't" group! Because fact is, that until you take into consideration irrational, wacky behavior caused by chasing EPC, the marginal cost of keeping the "Those I don't" affiliates in your program is minimal.
Now what about the marginal revenue half of the equation? Well, suppose 20% of the affiliates in the "Those I don't" group refer, say, 1 sale a year (Maybe they click on their own links to do their Christmas shopping!), and, suppose that TB has assigned 1,000 affiliates to this group. 20% x 1000 = 200 Sales from the unlucky group. If you don't consider epcmania, those are 200 extra sales that are just as profitable as those sales generated by the "Those I do" group, and 200 extra sales that would be gone if you had axed all those in "Those I don't" group. This is exactly the opposite of what TB should be doing to maximize their profit!
So why, oh why are affiliate managers losing sleep every time their EPC slips? According to profit maximization theory, a low EPC can be just as easily a sign of an outstanding program, with a smart profit maximizer at the helm! After all, if the cost of getting that one extra sale is almost nothing, why should TB care whether it took 5 clicks or 5000 clicks to generate it? The truth is, that TB shouldn't care!
But somebody cares.....and cares a lot, because it is not consistent with their profit maximizing strategy. To be continued....
1 (Notice I didn't say "affiliates who have no sales" or "affiliates who have been dead for 5 years"; even affiliates with sales are not safe from the axe if it took too many clicks to get them.)
|
March 29th, 2006, 02:50 AM
|
 |
Newbie
|
|
Join Date: March 5th, 2006
Posts: 266
|
|
Excellent post. Recently I was dropped from a famous Loans programme after only 3 weeks. I hadn't produced a sale and only brought 50 clicks. However, I have a strong earning history on CJ... so why dump me if I hadn't gotten rolling yet? Now they'll never know if I would've been an earner for them simply because they had to keep their vaulted top spot in the EPC sorting.
|
March 29th, 2006, 03:52 AM
|
The Smilie Master 
|
|
Join Date: January 18th, 2005
Posts: 5,559
|
|
Quote:
|
But, you say, there is a real marginal cost involved when sending email newsletters through the CJ system! True, and that is a real marginal cost! Please refer to my solution to problem 1; Don't email the "Those I don't" group! Because fact is, that until you take into consideration irrational, wacky behavior caused by chasing EPC, the marginal cost of keeping the "Those I don't" affiliates in your program is minimal.
|
So far I agree with you Jennifer, 
But, I will wait for part two!
Quote:
|
But somebody cares.....and cares a lot, because it is not consistent with their profit maximizing strategy. To be continued....
|
Sal.
__________________
.
|
March 29th, 2006, 07:58 AM
|
|
ABW Ambassador
|
|
Join Date: January 18th, 2005
Location: California
Posts: 2,068
|
|
Great post!
<~~~~~~waits with Mr. Sal
__________________
Mr. Mackin ~~~ mackin aaat web-traffic-management.com
|
March 29th, 2006, 08:05 AM
|
 |
general fuq
|
|
Join Date: January 18th, 2005
Location: Argieville
Posts: 1,320
|
|
Jennifer,
Some of what you've detailed here is the very essence of a lazy AM. If anything the "i don't group" is the group that should receive a newsletter based on how to improve conversion.
I agree the cost is minimal and the net result is branding if nothing else, but there is a tipping point. When an affiliate sends to much traffic that doesn't convert there will likely be a fault in either how they develop traffic or what they are preselling. This may result in a negative view of the advertiser by the visitor or a strain on system resources at the very least. I would think that the cost of the bandwidth is more than the price of an email to try to help that affiliate.
It's all about how active and intelligent the AM is...less active means more likely to axe... At some point you have to understand that not every affiliate will make you money even after working with them and the axe must fall.
2 cents spent
|
March 29th, 2006, 08:29 AM
|
 |
Canadienne extraordinaire
|
|
Join Date: August 11th, 2005
Location: Guelph, Ontario
Posts: 707
|
|
Yessssss! Jennifer, I'm so glad you posted this. It's the way we work with our affiliates on Share Results.
Thanks for putting it out there :-)
__________________
Susan Arts
Senior Vice President, Marketing
Moxy Media
|
March 29th, 2006, 09:19 AM
|
 |
Lite On The Do, Heavy On The Nuts
|
|
Join Date: January 18th, 2005
Location: Orlando, FL
Posts: 6,949
|
|
The reason is that many affiliates are not seasoned enough to understand things and they use EPC as their main criteria to find programs.
Average EPC is almost meaningless taken out of context of other factors in any program.
EPC ignores the individual affs traffic quality / relevance (hopefully your idea is better than average) and it completely ignores the costs of the traffic (whether ppc, placement, seo work to get ranked in a competitive niche, etc).
When affs stop focusing on it, so will those trying to recruit them.
|
March 29th, 2006, 09:25 AM
|
|
Lite On The Do, Heavy On The Nuts
|
|
Join Date: January 18th, 2005
Location: Orlando, FL
Posts: 6,949
|
|
And often, even when comparing 2 very similar merchants, EPC is meaningless as well.
I've seen the case where one merchant had a "get any traffic" mentality, so they had many unfocused aff partners bringing them traffic that didn't convert really well. They were a mature program that already had tons of great partners in house and they were seeking breadth. This merchant's EPC was much lower than their similar newbie competitor whose primary affiliates were trademark ppc poachers.
Merchants soaked with parasites can drive up their EPC as well - by moving their direct type-in traffic into the affiliate revenue bucket.
|
March 29th, 2006, 11:25 AM
|
 |
2005 Linkshare Golden Link Award Winner
|
|
Join Date: January 18th, 2005
Location: St Clair Shores MI.
Posts: 17,375
|
|
Do BHO's actual generate a click in CJ's system?
__________________
Webmaster's... Mike  and Charlie 
"What have you done today to put real value into a referral click...from a shoppers viewpoint!"
|
March 29th, 2006, 11:30 AM
|
 |
|
|
Join Date: January 18th, 2005
Location: Mansfield, TX
Posts: 10,961
|
|
Excellent! I agree wholeheartedly. "Average EPC" is usually a very poor metric, and a focus on that can cause affiliate managers to make some very stupid decisions.
But that begs the question... What would be a better alternative to an "Average EPC".
I proposed something similar to this to a network a year or two ago, but they never ran with it. Something like this would really help affiliates who produce determine what they can expect and how they rank.
1) For the time period you're analyzing, select only those affiliates with 2 or more orders. This will eliminate the absolute amateurs, the tire-kickers, and most of the affiliates who sign up just to place a personal order (with 100% conversion and a very high EPC). All of these can skew the numbers (inappropriately, IMHO). Make sure that this list contains at least 20 affiliates, otherwise it probably won't be enough to be statistically significant and you may need to sample a larger time period.
2) Sort these affiliates by total earnings and select the top 5-50%. You probably want no less than 10 and no more than 100. The higher earning affiliates will probably be more statistically relevant, while the lower earning ones excluded here may have some anomalies.
3) Sort this list by EPC and take the EPC numbers from the affiliates at the 10th, 50th, and 90th percentiles.
4) Round these numbers off so that they don't have artificial precision. An "EPC" of $30.71 isn't any more informative than an EPC of $30. I would round to the nearest whole number if it's under 10, the nearest 5 if it's 10-50, the nearest 10 if it's 50-100, and the nearest 25 if it's over 100.
I think the 10th and 90th percentile numbers would clearly indicate an EPC range that an affiliate could expect to earn in a program, and the 50th percentile number would be a much more reliable (and probably higher) EPC than a network-wide average. Low-performing affiliates would have a much small impact.
Myself, I would probably place much more importance on the 90th percentile number. That would be my target and that would be what I would strive to reach. If someone else can target their traffic well enough to get that EPC, I can too.
If any merchants would be willing to run their numbers and post their network-wide EPC and the 10th, 50th, and 90th percentile numbers from the above, I would be very interested in seeing how my theory works out in practice.
Last edited by MichaelColey : February 13th, 2007 at 02:08 PM.
Reason: Typo
|
March 29th, 2006, 12:22 PM
|
 |
Life is Supposed to be Fun!
|
|
Join Date: January 18th, 2005
Location: Los Angeles
Posts: 11,085
|
|
Excellent post Watchdog - thanks!
A few years ago I was axed by a merchant on CJ - After having sent them more than 5000 visitors in a month without a sale. LOL - I was annoyed with THEM for not closing my traffic but when they dumped ME, I had to find out why. The AM was gracious enough to let me know it was because my stats were messing up their EPC. I was and still am baffled ... this is a very targeted merchant, not one that could lead to erroneous pre-sell and thus, high volume clicks without sales. I should have dropped THEM for non-conversion. With my tail between my legs, I replaced all their links and gee ... the new merchant converted like crazy!
So, who's issue is it? Mine or theirs? Am I a "bad" affiliate because I sent so much non-converting traffic or are they a "bad" merchant for not converting all that traffic?!!
Chicken/egg theory anyone?
__________________
Peace,
Rexanne
Rexanne.com
Loving Everyone's Child Creates Magic
|
March 29th, 2006, 02:49 PM
|
 |
Super Sh!t Stirrer
|
|
Join Date: January 18th, 2005
Posts: 10,031
|
|
Stupid is as stupid does.
__________________
Comments are opinion unless otherwise noted. Remember, pillage first. Then burn. Half of all people in the world have IQs under 100. You best learn to trust ol' SSanf!
|
March 29th, 2006, 03:08 PM
|
|
2005 Linkshare Golden Link Award Winner
|
|
Join Date: January 18th, 2005
Location: St Clair Shores MI.
Posts: 17,375
|
|
Here's the real deal for AM's to ponder and learn. What is your sites average conversion ratio from ALL traffic sources?? (conversion ratio is number of clicks to produce one sale)
Then what is the average conversion ratio from your affiliate sales force?? Network # of sales divided into the # of clicks = conversion ratio. If you run a PPCSE/SEM campaign for the merchant you need to know that conversion ratio as well.
All the EPC crapolla was initially designed to never reveal to affiliates the average conversion ratio they could expect from sending targeted traffic. No existing network has the balls to publish affiliate enabled merchant's average conversion ratios. That important recruiting stats is left up to the AM's.
__________________
Webmaster's... Mike and Charlie
"What have you done today to put real value into a referral click...from a shoppers viewpoint!"
|
|
Discussions: 116,406 
Messages: 951,436 
Members: 61,011 
Newest: rolexwatch 
Online: 1338
Linear Mode
