The Perfect Affiliate Network Anti-Scrubbing Script

Yes, a very long title, indeed. Disclaimer: You’re not alone if you have to read this post 2-5 times to fully understand it.

After I posted about how I was split-testing networks, I immediately switch all of my traffic over to an affiliate network (for one particular offer).

However, after I switched, my conversions continuously dropped with this affiliate network.  I then switched all of my links up again to go direct with the advertiser.  I wouldn’t classify my current conversions as “great,” but they are “okay” and are converting about 5 times better than the second affiliate network I picked now.

This lead me to the idea of making a script that would rotate multiple offers but then automatically increase the percentage of redirection to a particular affiliate network (after finding which network converted the best).

Okay, let’s say after 1 week, we had the following conversions:

Network A: 5.5%
Network B: 6.0%

Now, that’s great. Our script would automatically redirect to network B (most of the time).

Alright, now our second week passes, and for this week we get the following conversion rates:

Network A: 5.5%
Network B: 5.0%

As you can see, after our script automatically decided to send most of the traffic to network B, network B decided to scrub us.  However, network A remained consistant–which obviously is where we should send most of our traffic.

How to determine which converts best on average (and automatically)?

You can automatically use confidence intervals to find which network should be used (and at what percentage).

If you don’t understand what I’m writing, play with this tool.

Say we have:

Network A: 20 leads, 200 clicks
Network B: 30 leads, 200 clicks

Calculating the confidence would give us:

That 93.4% would tell us to redirect to network B 93.4% of the time.  Statistically, it is impossible to ever reach 100% confidence (which is perfect for this script).

If conversion rates remained consistent, network B would eventually be used 99.99% of the time (since confidence would eventually reach 99.99%).  However, if conversion rates were to eventually switch to the favor of network A, the confidence would gradually lean towards network A, and network A would evidently be used most often.

In a future post, hopefully we can do some actual coding.