@CPMstars wrote in post #12
"Depending on your starting budget you need at least $200 to test properly, on an average of $0.2 click you will get 1000 clicks which should be enough for you to determine how profitable the campaign can be. You should set your benchmarks (CTRs and eCPAs) from the beginning and keep optimizing daily per source id and placement."
I don't disagree, but I have some questions. It seems to me that 200 bucks would be a bit modest (which likely is why CPMstars wrote at least $200) to test a campaign that had very many source ids and placements.
Suppose each placement only had 10 source ids and there were 10 placements per campaign.
That would sum up to 10 x 10 = 100. And 100 x 10 = 1000.
Given that the clicks were evenly distributed among the 10 source ids per 10 placements (unlikely), each source id would have only 10 data points, which I think would not yield enough info to make a decision to terminate for lack of performance.
And, the data gets even less statistically meaningful when the 10 data points per source id are divided among split testing a couple of different offers, with several images, with several corresponding headlines.
Here is my question: which really interests me because I cannot afford to lose more than 2 or 3 hundred dollars testing a campaign. Especially, if native is like PPC in that only about 1 in 5 or 6 will turn out the be winners.
How does one set up a native ad campaign to buy enough meaningful data for 2 to 3 hundred dollars? Is there a technique for narrowly focusing on source ids and placements? As for offers, images, and headline copy, I know there is native ad spy tools that can help with narrowing those variables.
"Depending on your starting budget you need at least $200 to test properly, on an average of $0.2 click you will get 1000 clicks which should be enough for you to determine how profitable the campaign can be. You should set your benchmarks (CTRs and eCPAs) from the beginning and keep optimizing daily per source id and placement."
I don't disagree, but I have some questions. It seems to me that 200 bucks would be a bit modest (which likely is why CPMstars wrote at least $200) to test a campaign that had very many source ids and placements.
Suppose each placement only had 10 source ids and there were 10 placements per campaign.
That would sum up to 10 x 10 = 100. And 100 x 10 = 1000.
Given that the clicks were evenly distributed among the 10 source ids per 10 placements (unlikely), each source id would have only 10 data points, which I think would not yield enough info to make a decision to terminate for lack of performance.
And, the data gets even less statistically meaningful when the 10 data points per source id are divided among split testing a couple of different offers, with several images, with several corresponding headlines.
Here is my question: which really interests me because I cannot afford to lose more than 2 or 3 hundred dollars testing a campaign. Especially, if native is like PPC in that only about 1 in 5 or 6 will turn out the be winners.
How does one set up a native ad campaign to buy enough meaningful data for 2 to 3 hundred dollars? Is there a technique for narrowly focusing on source ids and placements? As for offers, images, and headline copy, I know there is native ad spy tools that can help with narrowing those variables.