PDA

View Full Version : The fundamental problem with Commission Junction


Edwin
12-18-2001, 08:39 AM
(Warning: this is LONG - but if you use Commission Junction extensively you owe it to yourself to understand this!)

I like Commission Junction for a lot of reasons, notably its relatively simple interface (once you get used to it) and the range of merchants it offers, as well as the super one-cheque policy!

However, there is one really fundamental problem with CJ that I don't see people discussing much: the reliance on EPC numbers.

Now CJ defines EPC as "Earnings per 100 clicks" and presents those numbers over the last 3 months, and over 7 days, for people to select the advertisers and ad creatives that will perform best... or NOT!

Here's an example that illustrates the absurdity of this metric being used as the "fundamental unit" for comparing programs:-

Program A pays $100 per purchase of an oil-fired heater (this is a fictitious example by the way). About 0.5% of visitors clicking through the ad will become buyers so CJ measures an EPC of $50 (1 in every 200 click-throughs signs up, generating $100 commission, so the commission per 100 clicks is half that, i.e. $50) On the face of things, this looks VERY exciting! Wow, $50 per 100 clicks... (hold that thought for a moment!)

Program B pays $0.05 per free signup to a newsletter. About 20% of visitors clicking through the ad will sign up, so CJ measures an EPC of $1. (1 in 5 click-throughs sign up which means that 100 clicks will produce 20 signups and $1) Doesn't look so hot, right?

WRONG!

If you just worked through the logic (and this is the type of logic CJ *encourages* by the way it presents the global advertiser stats) you might have arrived at completely the wrong answer.

Why? Because EPC (earnings per 100 CLICKS) takes NO ACCOUNT WHATSOEVER of the "pulling power" of the ad itself i.e. how effective the ad is at getting people to click through!

Going back to our two examples, in Program A an average webmaster has to show 2,000 banners to get a click-through (hey, the purchase of an oil-fired heater isn't exactly top of many peoples' minds...)

Now remember, 100 clicks are worth $50 (we saw that earlier) so each click is worth $0.50. Now you can see that 1,000 banner displays are worth just $0.25 i.e. the program has a CPM equivalent of $0.25! Not looking so hot...

Let's look at program 2. A single click was worth $0.01 since it took 5 clicks to earn a measly $0.05. HOWEVER, people are being asked to sign up for a fun free newsletter, and the ad is pulling a 5% clickthrough rate. Suddenly, for every 1,000 banners being shown, there are 50 clicks... which means $0.50 CPM!

Frankly, these examples are just scratching the tip of the iceberg as in practice CJ spits up much more extreme cases.

Let me close by illustrating the above with some real stats from my site (of course, your mileage will vary but the PRINCIPLE remains the same!)

I ran the Half.com promotion and it pulled a $5.70 EPC (meaning for every 100 clicks I would generate $5.70). However, after showing 607,042 banners, I generated just *263* clicks (less than 0.05% clickthrough rate), netting me $15. This is a CPM equivalent of $0.02 (that's TWO CENTS CPM)

I ran a promotion for Streamload.com and it pulled a $0.09 EPC (meaning that for every 100 clicks I would generate just $0.09!!). However, after 5,542 impressions I generated 1,685 clicks (a 30.4% clickthrough rate) netting me $1.48 for a CPM equivalent of $0.27.

So the "headline numbers" ($5.70 EPC vs $0.09 EPC) - which is all CJ would give me to compare the two programs were it not for the fact that I ran both for a while to get more data - conceal the ALL-IMPORTANT DETAIL: Streamload outperforms Half.com on my site by close to 14x. That's 1400% better performance, to put it another way.

Until CJ decides to release EPM data for its campaigns, the stats it offers are frankly only useful for ONE very small thing: to see if the merchant is improving their program over time. If the 7-day EPC is higher than the 3-month EPC, then the merchant must have fine-tuned either the ads, the back-end sales process on their own site, or both.

Here's hoping you're now better equipped to make sense of CJ's soft-as-quicksand EPC reporting feature!

Edwin

gallery
12-18-2001, 09:15 AM
I understood this as well.

You are right Edwin.

rolli
12-18-2001, 10:06 AM
Hi,
and that seems to be the same with EPC rating.
I got a mail from a merchant, containing the following:

We regret to inform you that you have been dropped from the Playcentric - Music & Movies merchant program, and you will no longer be affiliated with this advertiser as of 12/10/01. This is not intended as disapproval of the quality or value of your Web site. Each CJ advertiser has the option to determine how many and what types of publishers they will approve to their affiliate program, and they may also make changes to their program which warrant removing some of their publishers.In this case, the advertiser has supplied the following reason for removal:

High number of hits and 0 sales; therefore hurting our C.J. EPC rating.



For me that only means, that I did my work very well, but the merchant's products and services are bad.

Jonny

SWD
12-18-2001, 10:30 AM
I agree in what you say!

jnestor
12-18-2001, 12:07 PM
You're right to a certain degree. EPC surely doesn't tell the whole story. It is a little more valuable than you suggest however. For example I think there are 3 or 4 sites that sell magazines at CJ. Comparing the EPC for these is constructive. I'd expect a "People Magazine" creative from all of them to get about the same CTR so the EPC could help you decide which merchant is the better choice.

The problems with EPC are as you describe (doesn't take CTR into account) and as rolli describes (encourages merchants to drop people with low EPC). However there are also problems with doing an effective CPM. Namely that the effective CPM will vary extremely widly based on the site.

What I really don't understand is why CJ doesn't allow the publisher to categorize their site and then calculate effective CPM for each advertiser within that category. Yes, it will still vary within the category but it shouldn't be to the same degree as across all sites.

The last problem with effective CPM and I think it's more subtle is that if you're not actually paid on a CPM basis, the effective CPM can and will change over time based on your traffic. If your site gets a lot of return visitors the effective CPM for any given advertiser will likely drop off over time. If you have a small amount of highly targetted traffic, you might get a great effective CPM. But getting a huge amount of additional traffic that's not targetted will surely drive that effective CPM into the ground.

LastActionHero
12-18-2001, 12:38 PM
The useful function of EPC is to compare ads within particular program. As Edwin pointed out between different programs it is not much of a use but within the same program the EPC might be safely be taken as a tool to compare the effective pulling power of that particular ad.

WildComputer
12-18-2001, 01:11 PM
The main value of EPC IMHO is it lets you see who has probably broken their tracking

If a merchant has EPC = $0.00 for 7 days and 3 months, don't expect to make too much money!

There is a merchant I know of that abruptly stopped tracking leads in mid January on the same day (I think) as they revised their web site.

Their EPC is still $0.00 for 7 days and 3 months, 11 months later.

jnestor
12-18-2001, 04:37 PM
LAH - I'm not sure I agree with you. You might actually see better overall performance with a banner that gets a lower EPC for a particular program. It goes back to the way that EPC is defined.

Let's say a program pays $1 per lead and has two banners.

Banner 1 says: don't bother clicking here unless you want to sign up for our program. You show 10,000 banners, it gets 10 clicks and 1 lead or $10 EPC.

Banner 2 says: come look at this awesome program that's free and win a playstation 2. You show 10,000 banners, it gets 1,000 clicks and 100 leads or $10 EPC.

You decide which is better. :)

Sure the numbers are made up and most likely exagerated but that's the problem with EPC. It alone doesn't distinguish between these two cases. In many ways the EPC measurement punishes banners that attract a lot of clicks. If the program has a broad interest and the site does a good job of selling it a better CTR and a lower EPC might make more.

ceo
12-19-2001, 05:28 AM
That was a good example.
A clear way for someone to understand.

I agree completely & with Edwin's thoughts too.
Most of us have already known about it but some incld. me have not actually calculated the difference. A large one it seems.

wly
12-19-2001, 06:19 AM
It's been only lately that I start to appreciate the value of the EPC.

LastActionHero
12-19-2001, 06:23 AM
Nicely put jnestor, but this is where the three month EPC and 7 day EPC comparision comes into focus.

In the above example the EPC for banner 1 will most likely be less than the EPC for banner 2 because banner 1 being of poor quality won't attract a consistent CTR and thus leads whereas banner 2 will perfrom strongly thus increasing it's 3 month EPC.

Robert from SI
12-19-2001, 09:16 AM
Edwin,

A fantastic real-world observation and explination.

demae
12-19-2001, 11:30 AM
> In this case, the advertiser has supplied the following
> reason for removal:
>
> High number of hits and 0 sales; therefore hurting our
> C.J. EPC rating.

Man... that advertiser's actions seem kind of like "cheating".

rolli
12-19-2001, 12:09 PM
The merchant brought this mail throug cj.com mail department, and I think cj.com knows of that. If they do not investigate in that, have no chance to know wether I was cheated or not.
Jonny

Kaiosama
12-19-2001, 03:12 PM
Originally posted by LastActionHero
Nicely put jnestor, but this is where the three month EPC and 7 day EPC comparision comes into focus.

In the above example the EPC for banner 1 will most likely be less than the EPC for banner 2 because banner 1 being of poor quality won't attract a consistent CTR and thus leads whereas banner 2 will perfrom strongly thus increasing it's 3 month EPC.

Umm low ctr doesn't equal low EPC... Just because an ad get's a lower ctr it doesn't necessarily mean it will get a lower epc. In my experience it happens to be the total opposite. A banner that doesn't entice anyone but those who really want to join a program usually get low ctrs and a higher epc.

Doesn't this discussion sound familiar? I'm surprised CJ didn't include an EPM or CPM.

kryton
12-19-2001, 03:29 PM
Edwin very good example.

Has anyone ever suggested EPM to CJ?

What Edwin said was the one the first things I realized once I understood all of CJ terminology. I've always looked at the program its self, and used the EPC to estimate an expected CPM.

EPC is useful because you can estimated your approximate CPM from it. Possibly (not always) more so than if you given an EPM metric. Here is my reasoning: all of CJ's ads are CPA, a host may not care about placing the ads in their best CTR locations. Or some may just stick random ads on their pages hoping to make just a bit more money, hence giving a low EPM. People who have sites with a very high page view to unique vistor ratio, may have been forced into CPA programs because of their low CPC. Or more generally people with low CTRs may have been forced into CPA ads networks. If people realise everything Edwin has explained then EPC can be a useful metric, but I actually believe CJ should use both EPM and EPC metrics.

FocaJohnL
12-19-2001, 04:24 PM
Edwin:

Great post! The EPC model *is* flawed if used as a sole measure of a merchant's performance, as you noted, but the problem is that there isn't a good way to measure effectiveness of a merchant and have it be true across the board. CPM is a different model, for example, which measures based on raw impressions, and people like to use that measure because the CPM rate divorces the conversion ratio from the effectiveness of the merchant to convert. The unfortunate problem with that is that people will experience widely divergent CPMs, so what might be a $50 CPM for some people will end up being a $.50 CPM for others.

Basically, most cost comparison models are flawed, and as an advertiser, it just pays to test, test, test, and generate your own numbers to determine what works for you.

Very good post, though, and very educational!

freakysid
12-19-2001, 06:57 PM
Great thread!

EPM - the problem here is that CJ can't really track this. It's up to you the webmaster to track how many views your banner/button/text link/ezine ad has had and work out your effective EPM from there.

Kaiosama
12-19-2001, 07:00 PM
Actually CJ can offer EPM as it already has the feature to let you see your own EPM. It only shows the EPC when looking at new ads/programs. Also CJ does track impressions as well (kinda required to get the EPM).

wly
12-19-2001, 10:26 PM
They should also publish CTR and the CR. :D IMO, the EPC is enough because CJ would like to brand themselves as a pay for performance network and CPM is not about performance. I think it is logical to go with merchants with the highest EPC within a category. The rest is up to the publisher to increase CTR, CR, etc. If EPC is high, it can speak of these diff. things:
a) Good CR
b) Good % commision
c) Good Cookie Duration
d) the merchant is not a dud

etc. etc.

Vrindavan
12-19-2001, 10:55 PM
EPC is the best data to compare
the ability to convert between different merchants of same category.

We do not get paid unless merchant can convert traffic to sales.
It is very disappointed to see that the bad merchants cannot convert sales for high quality traffic.

I will always pick the high EPC merchant to promote.
My commission go up fast.

CPM is useless to compare merchants' ability to convert.
And ths data is highly inaccurate.
1. inaccurate impression number tracking.
2. Most webmasters do not include impression code in their links (all text link, newsletter links, etc)

You may not agree with me.

joma
12-19-2001, 11:00 PM
Effective CPM would be an utterly meaningless statistic: Link type and placement vary dramatically from site to site, demographics also very dramatically, many CJ impressions go out in email newsletters which often can't be tracked, and a lot of sites do not include the tracking code which allows CJ to track impressions at all.

EPC is far from perfect, but it's a useful yardstick as long as you are aware of its limitations.

[Vrindavan just slipped in ahead of me saying much the same thing]


One thing I've noticed is that a merchant's vanilla text links usually have a dramatically higher EPC than banners and buttons (frequently, 10x or more). I *think* that's because webmasters doing pre-selling and designing custom links are using this linking code, but it's hard to know.

Kaiosama
12-19-2001, 11:01 PM
CPM shows you the return your getting over the exact amount exposure your giving the product/service. But CPM alone I believe would be a poor measurement for the full assessment of a merchant or link. Couple it with CTR or EPC and it would be great.

I don't see a good reason for cj to restrict us let us see just the EPC of the overall network stats. Newsletter links are a wholly different issue but I really don't see the reason for not including everything (tracking images etc) to make sure your link is being tracked properly on CJ. It only gives you valuable information that you can use to help you calculate what ads/services are best promoted on your site.

Leader
12-20-2001, 12:54 AM
The main thing I see wrong with EPM (as it pertains to me) is that I show multiple banners and text links on the same page, to the same merchant.

Since a visitor to the page is only going to click one of them (unless they're weird), that will mean that the unpopular ones will show a lousy EPM. It will also skew the merchant's EPM rating--again because a person is only going to click one ad on the page.

All the different metrics have their advantages and disadvantages. EPC is okay for showing the extremes, especially. I will admit to not thinking much of EPM because high-traffic, low clickthrough sites can make it seem way lower than it should be! A targeted sales page can drive in a huge EPM simply because only those interested will be seeing the ads in the first place. But the effect of sales sites on EPM is diluted by people with hundreds of thousands of un-or semi-targeted views/month or more!

CR is okay but it's hard to tell whose sales job is affecting it. Both the merchant's and the affiliate's sales job can highly affect the CR! Trying to find out who is responsible for a merchant's CR and EPC (both) can be a real detective job.

A combination of metrics should be employed, IMO, along with a knowledge of what your site's visitors are interested in, what price range they like, etc.

Vrindavan
12-20-2001, 01:38 AM
Kaiosama

1. We cannot put impression tracking code on text email.

2. The person can view stored email message more than one time or many many times later of the same message, the ad reponse is not the same as ads on website.

3. When we read email offline, impression tracking again is meaningless.

4. Someone has already point out, I do too, is that I will place more than one banner or text link in one webpage.
What about 30 or 100 product links from the same merchant in one webpage.

5. When a data is actually not accurate at all.
It will do more harm than good to us if available.
(wrong picture, wrong impression to actual fact)

Edwin
12-20-2001, 03:38 AM
Well, I pretty much covered my objections in the initial post, but lots of excellent points are being raised, so I thought I'd follow up briefly.

Remember: EPC is "earnings per 100 clicks".

So imagine you have a lead-based program that is converting MIRACULOUSLY well (i.e. every single click-ee becomes a user of the free widget service, or whatever is being advertised). This program is offering a payout of $0.05 per registered user. This in turns means that its EPC will NEVER go over $5, regardless of how good the program is, because 100 clicks x $0.05 = $5.

Now you have a second program that pays $50 per lead, but that converts 1:1000 on average. That will ALSO give an EPC of $5! (50/1000*100)

That's how murky the EPC measurement really is - it can show the same "headline number" for a program with 1000x the closing rate (ability to make sales) as another one.

You can at least figure out what the site's "closure rate" is across all visitors sent to it by dividing the EPC by the price per lead giving you a closure rate as a %age.

For example, if a given sponsor program has an EPC of $10 and a price per lead of $0.50, then you can see that it has a 20% closure rate (the sponsor converts 1 in 5 of all visitors into "buyers").

Beyond the above figure, what is needed is a totally new measurement, the CPCPM number, which shows how many clicks are generated by 1000 "impressions" of the ad across the whole network.

This CPCPM could then be tied in with the EPC to give an *approximate* per-CPM value. Doesn't matter how far off the real number this is, as long as ALL calculations for ALL advertisers are made in the same way - that way, the errors basically blend into a foggy grey mess, leaving "comparable" numbers that people can use to make real decisions.

I believe, frankly, that CJ is DELIBERATELY OBFUSCATING this kind of tracking on a network-wide basis, since greater transparency in its up-front reporting would have the inevitable consequence of leaving hundreds if not thousands of sponsors with no affiliates at all, once the full horrendousness of the advertising material they provided has sunk in...

For example:

A merchant pays $40 for each "FREE" mobile phone registration, with an EPC of $20 for a given ad. This means that out of every 200 visitors you send them, 1 signs up (on average).

BUT here's what CJ doesn't show you until you waste hundreds of thousands of impressions testing... in this example, the merchant's advertising material (ON AVERAGE - we're *always* talking averages!) draws a 0.05% click through rate. That means that for every 5,000 banners/text links shown, 1 visitor goes to the site.

Suddenly, you can see that it takes 5,000 impressions to get a visitor and 200 visitors to close a sale... in other words, you have to show 1,000,000 ads (of whatever kind we're talking about) to make 1 sale! And that sale just brought you $40, giving you an effective CPM of $0.04!

But you will NEVER find that kind of number out from the CJ data until AFTER you wasted your 1,000,000 impressions (or gave up earlier in disgust!)

So in summary, the ONLY thing EPC is good for is as a basis to understanding how effectively a merchant converts visitors into buyers.

If they have a good closure rate (e.g. 1:10-1:100 depending on the product in question) it may be worth doing a quick test (e.g. 10,000 impressions of the ad material of your choice) and seeing what happens.

For example, if you choose a specific TEXT link that is the best-performing EPC link the merchant offers in its advertising arsenal, and your quick burst of mental arithmetic (EPC and CPA) shows that it converts 1:10 visitors into buyers, then if you've sent 30 visitors you can be pretty sure it doesn't work well for your site's audience. And if after 10,000 impressions, you only sent 1 visitor, well you also know something very valuable about how that merchant will perform for you.

Edwin
12-20-2001, 03:51 AM
I know I wrote a lot above, but I had a few more things to add:-

A) If you have the choice between a high per-ticket program and a low per-ticket program (e.g. $50 per sale vs $5 per sale) with the same EPC (and the same target audience) then the $5 per sale program looks more appealing than the $50 per sale.

Why? Well, I can see that the merchant paying $5 per sale is 10x (TEN TIMES) as efficient at closing a sale with the average visitor as the merchant paying $50 a sale. So even if my site's traffic is below "average" in terms of quality and susceptibility to purchasing, I'm still likely to be able to close at least SOME sales and see some revenue coming in... even if my own EPC turns out to be lower, at least it won't be ZERO.

B) What frightens me is seeing VERY extreme EPC swings within a single merchant's range of ads. Remember, EPC is measuring what happens on a merchant's site, NOT on your site (the only contribution your site makes to EPC is to send 100 visitors)

So if a merchant has Banner A with a $5 EPC and Banner B with a $0.50 EPC, that sets off alarm bells in my head. Granted, there's also the opportunity to try and analyse WHY one ad may pre-dispose people to buying 10x more than the other ad, but more likely there is some flakiness of their back-end tracking system... otherwise why would 100 people who came through Banner A from a site be 10 TIMES as likely to buy as 100 people who came through Banner B.

The only exception to this is if one of the ads mentions a specific price on it, or other similar limiting conditions. If I see "Cheap Playstation 2 games" or "Playstation 2 games - just $18.99!" my reaction will be totally different. Perhaps people clicking on the "Cheap Playstation 2 games" banner believed (on average) that "Cheap" meant "Under $10" and hence were put off buying many games.

ASIDE: By carefully combining targeted text links and selecting CJ programs that matched my site's audience precisely, I have increased CJ revenue on my site from about $300 a month to over $1,500 for December (if current trends hold true). That's with NO increase in traffic at all, just hard work at the picking ads and testing stage.

So if you have a well-trafficked site, the effort you put in really CAN be rewarded, as long as you take care to always understand (and remember or note down somewhere) everything you test and all the results you get.

Good luck!

Edwin
12-20-2001, 04:19 AM
I'd just like to add (yes again, groan!) a few REAL examples taken from CJ's advertiser list. I haven't tried these advertisers, it's just math - but VERY educational math nonetheless.

A) Premier Equity is paying $25 per lead to its Home Equity products and has a 7-day EPC of $57.13. This means that for every 43.75 (let's call it 44!) visitors *their average affiliate* is sending them, they're closing one "lead".

B) InsuranceQuote Services is paying $16 per lead and has a 7-day EPC of $43.97. This means that for every 36.4 (37) visitors sent to their site, 1 lead is being generated!

C) NationalPayday (cash advances) is paying $11 per lead and has a 7-day EPC of $1.78, meaning 1 lead for every 618 visitors.

D) ATT Wireless is paying $35 per Sale and has a 7-day EPC of $0.36, meaning 1 lead for every 9,722 visitors (!)

FINAL CLOSING HINT: If you look at the bottom of the "Advertising list" tab on the "Get Links" page within CJ, you'll have access to the ULTIMATE WEAPON: you can download ALL the data for ALL their advertisers in Excel-ready CSV format! After that, it's easy to do any number of formulas on the data to find useful trends and patterns...

wly
12-20-2001, 04:37 AM
Yup the name of the game is conversion. These way both side wins. We would definitely work with merchants with high EPC because we know they can close the sale. Merchants would also like to work with high EPC affiliates because they can deliver the sales.


Just to illustrate, for example if eBay drops its cookie duration to 1 day, I can imagine their EPC sinking to stinky waters. Your CR will definitely drop like s*** (no pun intended) thus affecting your EPC. If you take a look at eBay, their EPC has been slowly dropping the past few months because they have implemented many changes. EPC is a fast way of determing the total package of the merchant.

Don't get me wrong I used to adhere to the EPM policy should be implemented but now I came to realize that EPM only works well in determining the CR of independent banners but not the merchant in general. It would be safe to bet that those high CTR trick banners will have a low EPC because they have a very CTR but poor CR but hey you can earn more with those banners becuase it gets clicked more often than those other lousy banners which have pathetic CTR. I guess its safe to say that most banners in general can have a high EPC because nobody clicks on banners in the first place and if you can get a pretty decent banner CTR most probably 99% of the time people will not buy thus a pretty low EPC. So I can conclude that if you see banners with high EPC, chances are nobody clicks on those banners and if the EPC is low chances are you can make something out of it but not much.

And as Leader pointed out I sometimes use different banners on the same page. I use 468, buttons and text links and sometimes I use different adverts for the same merchant. This would really ***** the data for EPM because CJ tracks impressions of diff. banners seprately. BTW, nice seeing you, Leader of CJ affiliates, posting here. ;)

So to conclude I think this is really a very planned move by CJ and not something thrown up overnight. Of all networks, nobody comes close to CJ in the eyes of affiliates, maybe Linkshare sometime soon if they aggregate their checks.

wly
12-20-2001, 05:07 AM
Edwin,

I think I would go with the high ticket merchant. Granted that both EPC are the same, I think you can have a better chance of earning more because your EPC can be maxed at $50. There are many factors that affect the EPC. I will look closely into it though as to why they have a very low EPC, maybe a 90% reversal rate? This is one area I would look closely into before considering a merchant not to mention cookies.

Edwin
12-20-2001, 05:40 AM
What you just said contradicted the point of all my articles!

The merchant with the high ticket price will be MUCH MUCH HARDER to test on your site, since you're going to have to send a massive number of clicks in order to find out if your particular audience can sustain the same EPC as "average"

The low ticket price merchant is very easy to test... just send 3x as many clicks as their average "conversion factor" and you'll know whether their program fits your site's audience.

(the above assume the case where the two merchants' EPC figures are IDENTICAL)

Trust me, ON AVERAGE you're not going to "get lucky" with the $50 a lead merchant, because the EPC figure is already reflecting all the people who got lucky and all the people who got nothing.

The ONLY time I would contradict myself is if you have very low traffic, so that you would be unlikely to make the minimum check level through accumulating low ticket sales. Then you're basically playing "lottery" with the program: keep sending traffic in the hope of beating those odds...

Edwin
12-20-2001, 05:45 AM
You said:-

Yup the name of the game is conversion. These way both side wins. We would definitely work with merchants with high EPC because we know they can close the sale. Merchants would also like to work with high EPC affiliates because they can deliver the sales.

This is wrong!

Actually, as I explained at some length, EPC says NOTHING AT ALL about the ability to convert sales.

You have to take the EPC figure and DIVIDE it by the payment per lead to calculate the ability to close sales!

E.g. EPC of $50 and payment per sale of $2 means 25 sales per 100 clicks (1:4 closure rate)

E.g EPC of $50 and payment per sale of $200 means 0.25 sales per 100 clicks (1:400 closure rate)

The EPCs are IDENTICAL but the closure rates are totally different.

wly
12-20-2001, 09:09 AM
Sorry for being vague. This is all just my opinion and I'm in no way saying your wrong.

IMO, EPC does measure CR but not CR only. It is only part of the total package. There are many things that affect the EPC.

a) cookie duration (if a merchant has a 1 day cookie compared to another with 45 days then chances are the 45 day cookie merchant will have better CR).

b) reversal rate (this affects EPC as well)

c)% commission on a product ($5/lead can get you a max of EPC=$500 & $50/lead can get you a max EPC of $5000)

In the end, EPC measures the overall perfomance of a merchant with other merchants in the same category. I should have written performance in my previous post and not conversion.


As for the second post of the $50/lead, it was really just a matter of choice for me. If everything is constant i.e. (both merchants have same 3month and 7day EPC, same CTR, you drive the same targetted traffic.) In the end you're earnings should also be the same per 100 clicks.

What we see are network averages. Some might do much better with the $50/lead merchant (i.e. convert 1 sale every 100 clicks) Or you might see after 1000 clicks then bang 3 sales. Affiliate marketing has evolved and it's not as easy as CPC where your earnings per click is constant.

ceo
12-20-2001, 09:48 AM
Originally posted by Edwin
This is wrong!
Actually, as I explained at some length, EPC says NOTHING AT ALL about the ability to convert sales.
You have to take the EPC figure and DIVIDE it by the payment per lead to calculate the ability to close sales!
E.g. EPC of $50 and payment per sale of $2 means 25 sales per 100 clicks (1:4 closure rate)
E.g EPC of $50 and payment per sale of $200 means 0.25 sales per 100 clicks (1:400 closure rate)
The EPCs are IDENTICAL but the closure rates are totally different.

I think that gives a very clear understanding. With all the reams of notes you make I just can't read all that & figure it quick ; )
That short & simple example did it for me.

Anyone think that is wrong ?

joma
12-20-2001, 11:47 AM
Edwin, I agree with most of what you're saying, but here are responses to a couple of your points:


1. I don't agree that a low-ticket merchant is necessarily better than a high-ticket one with the same EPC, or even that a high-ticket one is harder to test. Maybe the higher-ticket merchant will do better, so why avoid it? It doesn't take long until you see that the clickthrough rate is 1 in 100 or 1 in 10,000 or whatever.

I think what it comes down to is that you have to understand your site and your audience, then stop fighting them and give them what they want. If I went with all the highest-EPC programs, my site would be plastered with credit card and term life insurance offers (and I'd be starving). Yet I have one very lucrative sponsor with a low EPC, and which I guess wouldn't do well on most people's sites.


2. I'm not concerned about the EPC variation between different ads for the same merchant. As I hinted at earlier in this thread, people who just slap up a button or (especially) a banner will get a poor clickthrough rate and those clicks will probably not be pre-sold or well-targeted.

On the other hand, superaffiliates who take the time to find a product well-suited to their audience, and integrate an ad naturally into their site, will tend to customize the text links. And when done well, that approach can vastly increase an ad's conversion rate.

Kaiosama
12-20-2001, 03:23 PM
Originally posted by Vrindavan
Kaiosama

1. We cannot put impression tracking code on text email.

2. The person can view stored email message more than one time or many many times later of the same message, the ad reponse is not the same as ads on website.

3. When we read email offline, impression tracking again is meaningless.

4. Someone has already point out, I do too, is that I will place more than one banner or text link in one webpage.
What about 30 or 100 product links from the same merchant in one webpage.

5. When a data is actually not accurate at all.
It will do more harm than good to us if available.
(wrong picture, wrong impression to actual fact)

I should have made myself more clear... I mean that newsletters are a totally different issue but for webpages i really don't see why one would exclude tracking images. I should have said webpages specifically. I'm not even asking CJ to record impressions for newsletters and if possible they could have separate newsletter links.

About point number four yes it's entirely possible to do that but once we visit the individual links we can see the epm of each and judge an advertiser using that.

CJ can also do a number of things to avoid the product issue like not include it in the overall EPM of an advertiser. Yes, that's not a perfect solution but it's a start. The thing is that with EPM or even CTR you can calculate alot of important data before trying out a merchant.

You can place an ad out there with a great epc but get a very poor CTR. I'm not saying that EPC doesn't help but wouldn't the CTR or EPM also help? Wouldn't it help to know what the average CTR of a certain ad has? Fineclicks.com seem to believe so and it doesn't seem to be hurting them at all. Getting the EPM would be very simple once you know both the EPC and the CTR.

Edit: Don't ask me why i keep on writing ebay instead of CJ, I'll try to re-read what I write next time.

wly
12-20-2001, 06:02 PM
Just to illustrate further, this is what I'll do when I'm faced with this dilemna.

First thing off, it would not be good to just assume that because you see that both merchant's EPC are the same that they are the same. I would definitely do further investigations as to why the high ticket merchant has a low EPC.

a) check to see if they have a high reversal rate (80% or more?) maybe this explains the low EPC.

b) cookie duration is one day?

If everything seems fine, I would definitely check the individual links. Just a scenario:

Merchant's A ($5/lead) text links has an EPC of 1.00 while banners on average has an EPC of 5.0 but the whole average for this merchant is 2.0. Merchant's B($50/lead) text links has an EPC of $10 while its banners has an EPC of 0.30 (due to very high CTR). Both might have the same EPC but clearly you can see the winner.

Vrindavan
12-20-2001, 10:42 PM
Edwin

Actually, as I explained at some length, EPC says NOTHING AT ALL about the ability to convert sales.

You have to take the EPC figure and DIVIDE it by the payment per lead to calculate the ability to close sales!

E.g. EPC of $50 and payment per sale of $2 means 25 sales per 100 clicks (1:4 closure rate)

E.g EPC of $50 and payment per sale of $200 means 0.25 sales per 100 clicks (1:400 closure rate)


We are comparing EPC to similar merchants.
You are giving examples like
1. selling magazine $2 per sale
2. selling a merchant account $200 per sale

Leader
12-21-2001, 02:16 AM
Joma's got good points. I will second them and save myself some typing since I'd only say the same thing, in that regard. I can also understand the comment about credit cards and loans--their high EPCs haven't helped them make money on my sites either.

I will add this, though:

If 2 merchants are selling the same thing at the same price, and paying the same commission, but one has a huge EPC and one has a lousy EPC, I think it can be safely said that something's wrong with the low EPC one and that the EPC is a fairly good red flag. Finding out what's wrong, and whether it can be overcome from the affiliate side, takes a bit more work!

Other than that, I can see the flaws in EPC that Edwin has pointed out as to the EPC not showing the full picture because it doesn't take the CTR into account.

Wly, thanks for the welcome!

rolli
12-21-2001, 10:05 AM
Hi,
all we are discussinghere is more theory. But if you really want to understand what we are talking about:
login to your cj.com account
goto the run reports tab
goto the performance report
click on 50 best performing links/ last year


The datas you will get explain everything.

Jonny