Well, now that all depends, does it.

First off, your sponsor is in the business of making money, just like you are. If in your mind (him making money) = (fucking you over), then yes. Obviously. If you find yourself thinking along those lines, maybe a more important question for you is should I be my own sponsor ?

The easiest way to see if your sponsor is fucking you over is to compare with another sponsor. Should the other sponsor do better, it doesn’t actually follow the first one was fucking you over, but the difference is more semantic than financial.

And so we arrive at the split test, aka the A/B test. They live close to the heart of internet marketing and are dearly loved by all affiliates (that make any money).

To implement one, you will need two sponsor codes, and a bit of php, like so :

<?php

$ads = array(’http://sponsor1.com/?aff=jbloe’,'http://sponsor2.com/?aff=jbloe’);

$value = mt_rand(0,1);

echo $ads[$value];
?>

Simple enough, but lets go through it. The first line loads an array, called ads, with the two sponsor codes you have. The 2nd line generates a random value between 0 and 1 (by using the Mersene Twister , a very good random number generator). The last line simply outputs one of the aff codes.

So you run this for a week, during which week say 20,000 people saw the ads. We know pretty much as fact that 10,000 people saw the ad of each sponsor, then.

Now suppose sponsor B sees 1478 clicks, and sponsor A 625 clicks.

Is A fucking you over ? Possibly. Before starting the drama, consider a few points.

Are both A and B reporting the same thing ? Some sponsors report raw clicks, some sponsors report unique clicks. If A is reporting uniques and B raws, then maybe A isn’t fucking you over after all.

But let’s say they both count the same thing, uniques. How can one have 6xx and the other more than twice that ?

Some sponsors filter certain clicks, such as proxies, web browsers that don’t accept cookies, certain locales, etc.

For instance, if your traffic consists of 10% proxy, 5% nocookie, 25% India 35% China, and program A reports filtered uniques, whereas program B reports unfiltered uniques, there’s no surprise A’s figure is 60% lower than B’s.

You can always ask a sponsor if they show uniques or raws, and generally they’ll be happy to answer, but if you ask them about filtering and such, especially as a freshly signed-up webmaster they don’t know, it’s possible they will feel uncomfortable and not tell you.

All this is not too important anyway, what really matters are conversions, since that’s where the money is.

Suppose your stats look like this :

A : 625 uniques, 0 conversions 0: 625 ratio;
B : 1478 uniques, 0 conversions 0:1478 ratio;

At this point you have two options. Either research and try to understand the global sponsor conspiracy against affiliates, with the goal of eventually exposing it for the filthy degenerate fraudulent scheme that it is, or else research and try to understand how to actually generate traffic that’s worth something.

If you incline towards that 2nd option, a good place to start’d be understanding your traffic.

But say your stats are more like

A : 625 uniques, 7 conversions 1: 89 ratio;
B : 1478 uniques, 1 conversions 1:1478 ratio;

Logically, it would seem, if your traffic is worth 7 conversions to A, who gets less than 1/3 of it, then obviously B is jipping you to the tune of 10-20 conversions per period. Right ?

Not really. There’s a few factors that could account for the difference :

I. Ads. If A’s ad reads “click here for very expensive stuff” and B’s ad reads “click here for mostly free stuff” then obviously fewer people will click on A’s ads, but those who do will be more inclined to spend money.

II. Niche. If your website mostly has pictures of Marilyn Monroe, and sponsor A sells BDSM whereas sponsor B sells Babes, the end result should not terribly surprise you.

Any website has a core slice of traffic that are people familiar with it specifically, at least to some degree (let’s call them C-visitors), and a larger chunk of passer-by type of traffic, that are maybe interested in the general theme (let’s call them P-visitors).

Now if you consider you have 2000 C’s and 18000 P’s, to make your original 20k, let’s see a scenario.

1000 C’s get exposed to the BDSM ad, and 615 go for it. Another 1000 C’s get exposed to the babe ad, and 19 go for it.

10000 P’s get exposed to the BDSM ad, and 10 go for it. Another 10000 P’s get exposed to the babe ad, and 1469 go for it.

Apparently, A has managed to attract 1/3 of your outbound traffic, and B 2/3 of it, but in fact A has attracted 60%+ of your core traffic, and virtually none of the passer-by, whereas B has done pretty decently with the passer-by traffic, but failed miserably on the core traffic.

Maybe it’s time you read up on pass-through sites vs bookmark sites, because you’ve made a bit of a mess of things.

III. Pure chance. It IS possible that it just so happened that B’s add was showed to all the people inclined to click but not buy, whereas A’s add was showed to all people inclined to click and buy. By pure chance.

To estimate how likely something like this is, you will need to understand something called the Poisson distribution. It is a very useful bit of math that tells you what are the odds that an event happening with a given probability will in fact happen in a given interval.

For instance, if your traffic converts 1:200 on average, what are the odds it will not convert once in 2000 clicks ?

In its full gory, the function is written as

but it’s nowhere near as complicated as it seems. Lambda (λ) equals your conversion rate, so if you average 1:200 and want to know the odds you will convert over 2k clicks, λ = 10. If you convert 1:150 and want to know the odds to convers over 15 clicks, λ = 0.1. Etc.

K is the number of occurences.

So now let’s apply. The odds of you doing 0 conversions over 2000 clicks, when your average conversion is 1:200, is

(10 ^ 0 ) * e ^ -10 / 1 = 1 * 0.00045 / 1 = 0.045%

You raise λ to the power of the number of occurences. Since this is 0, we get 10 ^ 0 which is always 1. You raise e (which is 2.71828, the root of natural logarithms) to the power of -λ, you multiply these two and you divide the result by the factorial of the number of occurences.

A factorial is multiplying all the integers smaller or equal. By definition 0! (zero factorial) equals 1. So, 3! = 3×2x1 = 6 and 5! = 5×4x3×2x1 = 120.

Now let’s apply this to see what are your odds to get exactly 1 conversion in 2k clicks with a 1:200 average ratio.

(10 ^ 1) * (e ^ -10) / 1 = 10 * 0.0045 / 1 = 0.45%.

What about 2 conversions ?

(10 ^ 2) * (e ^ -10) / 2 = 100 * 0.0045 / 2 = 2.25%.

What about 5 ?

(10 ^ 5) * (e ^ -10) / 120 = 10000 * 0.0045 / 120 = 37.5%.

If you’re a hard working, serious affiliate, like I know you are, you’re going to calculate for yourself the 3 and 4 conversion cases that I skipped.

Now what about the other example. What are the odds to get 0 conversions in 15 clicks, if you average 1:150 ratios ?

λ here is 0.1 (15 / 150), so the formula goes

(0.1 ^ 0) * (e ^ -0.1) / 1 = 1 * 0.9048 / 1 = 90.48%.

What about 5 ?

(0.1 ^ 5) * (e ^ -0.1) / 120 = 0.00001 * 0.9048 / 120 = 0.00007%.

And if you’re a hard working, serious affiliate, like I know you are, you’re going to tell me what are the odds to get 10 sales in 10 clicks, if you convert 1:100 on average. In exchange, I’ll tell you how to calculate the blasted e ^ x business on a windows system :

Go to the calculator, click view and select scientific. Then copy to clipboard the value of e from here : 2.71828182845904523536 and paste it there (edit > paste) then click the x^y button, then click the ( button, then type in -x, your number, then click the ) button and the = button. Voila.