Split Testing Offers With Highly Different Payouts (5)
So I recently set up a new campaign split testing 2 offers, one of them over 2 different networks, so 3 total. The payouts are as follows:
Offer 1, Network A: $ 8.0
Offer 1, Network B: $ 8.5
Offer 2: $ 2.0
Right now, traffic is split evenly among the 3. As the payout of Offer 2 is way lower (conversion rate is substantially higher as suspected), would it be a good idea to send more traffic to the higher paying ones to account for the big gap, sth like 40 : 40: 20?
Nope. Run them evenly and get comparable data. The EPCs are all you care about.
The fear here is that you're likely going to be wanting to make decisions with a sample size that is already too small, and exaggerating that position can make you do some silly things.
What you'll want to do is this:
Offer 1 Network A: 25%
Offer 1 Network B: 25%
Offer 2: 50%
So you don't take the payout into consideration at all?
The general consensus is that the higher the offer payout, the higher your sample size needs to be (drastic example: CPL vs CPS). That's why I thought about distributing more traffic to Offer 1.
Having the same sample size for all offers makes sense too though, so I'm a bit confused 
I don't think it's going to matter too much unless you a) need data NOWNOWNOW or b) are dealing with a really drastic difference in payouts - 10-1 or more. Either approach should work.
Provided you stick to solid statistical conversion rules (which I'm pretty sure you will 
) and use reasonable sample sizes, how fast you gather the data shouldn't matter too much - although obviously you'll want to gather it over 24 hours or so just to eliminate weird variance factors.
Personally I'd probably run the traffic evenly, but that's just because I'm lazy and traffic balancing is a pain in the ass if it goes wrong
I'd go with the advice above. Very good point about the urge to make decisions with a small sample, especially if you see the high payout offer converting early on. Temptation is always to axe the lower payout offer.
Split your distribution of networks evenly, but send equal traffic to each offer.
You could argue that even once you find that an offer converts on one network better than another, it still makes sense to split the traffic (as long as cashflow isn't a problem). You're less likely to get shitcanned from multiple networks at once if quality is weak.
Home > General > Affiliate Marketing Forum