Split Testing The CPA Networks (6)
Heyyy, are there anybody good with statistics and could suggest a way to split-test the networks themselves? For example, you test a few offers from the network A and a few offers from the network B. Is there a way to calculate a statistical significance so you can say that there is a 90% confidence that network A is better than the network B in terms of either ROI or profit?
If not, at what point do you decide to not work with a particular network because there is a better one?
Originally Posted by mantas
I've found some moderate variance doing that. I think it just comes down to what kind of tracking setup each one has and how closely they watch for leads that don't fire, etc.
I don't think there'd be any way to test out one network against another as a whole though, other than just subjectively based on your experience with the various individuals offers.
I guess you're looking to avoid conversion scraping as much as possible.
The best way would be to split-test the same exact offer from 2 or more networks. But you have to keep present that you need a considerable amount of data to do that. Besides this, you would need to take into account the payout value. It tends to be different from network to network.
Sometimes a slightly lower conversion rate with a higher payout can become more profitable than if you get paid less with a slightly higher cvr. Make sure to take into account all differences between the networks.
Sometimes a slightly lower conversion rate with a higher payout can become more profitable than if you get paid less with a slightly higher cvr.
Payout is one thing but CVR is just as important.
The highest payouts won´t help at all when it just doesn´t convert.
For example there are networks that offer highest payouts for offers so that it looks very attractive but then they have to shave/scrub to don´t run the offers on loss.
On the other hand you can have networks that offer lower payouts compared to others but run the offers without any "strange" tactics so that the higher CVR results in better performance for the affiliates.
Anyway, about the original question.
In the end the only thing you can do is to grab few similar (or even better same) offers from the networks you want to test and then just splittest.
But as @jack_l already said, it can also be different from offer to offer.
In my opinion it´s more important to stick with well-known and reliable networks, especially in the beginning.
Test few networks, when a network isn´t working at all for you move on and don´t try to make the impossible possible.
It´s just the way it is, some networks work very good for others but don´t work at all for me and vice versa.
When it´s not converting for me I just move on.
But there are also networks that don´t work for vertical A but work good for vertical B.
Another important thing to keep present is the payment terms. Especially when one sits on a tight budget and needs to improve the cash flow.
Say you’re running the same identical offer from two networks.
Network A pays every 15 days, while network B every 30 days.
If the combination of payout and cvr difference can potentially yield an extra 5%-10% in terms of ROI, but you’re on a tight budget, then you’d better stick with network A until you can support longer payout terms.
There are indeed a lot of variables to be taken in consideration.
Hi. Let me tell you how it's done in gambling. Usually affiliates test out networks the following way: they use the same app to promote multiple offers from CPA netowrks, separate them using deeplinks and simultaneously launch them for the same GEO and the same audience. After that, they identify the one that suits them best.
Home > General > Affiliate Marketing Forum