When it comes to affiliates, one of the most common questions I get asked is how do I know I can trust my affiliate partner? With affiliate companies popping up every single day (and closing down every single day), how do I know that the company isn’t shaving conversions or doing some other shady stuff?
Sadly, there is just one perfect way to confirm conversions and that’s to add tracking to the final success page (the page a visitor reaches once they complete the necessary action) and getting that is nearly impossible for most major corporations. Would Citi let you add a pixel on their success page? No. A smaller tech company, without all the levels of red tape, might but for many companies it’s not worth the hassle.
That said, you can confirm clicks by simply tracking actions on your side. You can do this by setting up event tracking in Google Analytics. If you have Analytics installed, it’s a simple matter of adding code to the OnClick event handler.
Split Test Affiliate Companies
The only way to know if an affiliate company isn’t shaving leads is by running a split test. Send half your traffic to one affiliate company, half to another. If the conversion rates deviate too much, you know who to trust. Sometimes companies can offer higher CPAs because they’re shaving leads, so the effective CPA is much lower. The only way to confirm this is by testing this way.
Be sure to run the test long enough to be confident that your data is statistically significant. If one company converts 10 out of 100 and another converts 15 out of 100, that’s a difference of 50% but it’s really only 5 people. Is that significant?
What is Significance?
The best explanation I found, which includes the math behind it, is here. Jason Cohen uses a Pearson chi-square test to calculate whether a difference is statistically significant.
Essentially the equation is:
N = number of trials, a trial is a conversion – NOT a visit
D = half the difference between winner and loser (clicks)
Significant if D2 is larger than N
For our simple examples of Company A converting 10 on 100 visits and Company B converting 15 on 100 visits, we calculate the variables to be:
N = 25
D = (15 – 10) / 2 = 2.5
D2 = 6.25
D2 is not greater than N so it is not statistically significant.
Multiple each number by a ten and what do you get?
N = 250
D = (150 – 100) / 2 = 25
D2 = 625
D2 is greater than N so it is statistically significant.
In some cases, it doesn’t take long to get to the truth. In other cases, it can take a while.
This is, of course, a win win situation. If you find out one company is statistically better, go with them after you factor in payouts. If you don’t find out either is better and both pay the same, you know have a suitable backup that you’ve vetted in case one company’s servers go down.
Lastly, remember that this does nothing to explain why one company performs better, just that one does (or doesn’t).
As President Reagan said – “Trust but verify.”
How do you verify affiliate companies?