When To Cut Offers When Optimizing? (6)
Hey guys so lets say I am starting a campaign and start off with something like 20 banners x 3 LPs x 3 offers
During the optimization process, roughly how many clicks would you say you need until you feel comfortable to narrow it down to just 1 offer?
500 clicks? 1000 clicks? (overall total banner clicks, not LP clicks)
Thank you in advance.
Hi, jabong82!
I’d like to ask you some questions just so that I can give you the help you need. You’re using 20 banners and 3 different landing pages. Are you only on one spot or several? Are you using a DSP or an Ad network? Are the offers in a rotator? Which offers are you promoting?
Why do I ask? Because – as far as banner optimization is concerned - I think you’re using too many banners and landing pages in case you’re only playing on one spot. This means you’re splitting your traffic among all these banners and therefore it’s gonna take more time for you to gather the info you need to make the right decisions. In case you’re only on one spot, I advise you to use only 3 banners. You shouldn’t use more than 5 (in a very high volume country).
When optimizing, you need enough stats in order to make solid decisions, but you can’t base those decisions on clicks alone. It’s usually necessary to let the campaign run for a few days in order to get enough stats. How long has the campaign been running? What’s the segment? The strategy for countries with a low volume (and generally high payouts) is completely different when compared to the strategy for high volumes (and generally low payouts).
In general - when optimizing - you have to compare the CPA with your offer’s payout. As a rule, if the CPA (of an offer, device, browser, OS, etc.) is already two times higher than the payout, then it’s very hard to make it profitable and you can safely cut it. If it has never converted and - therefore - you don’t have the CPA, then you compare the cost with the payout and you follow the same rule.
There are other useful parameters such as the CTR and the CVR. If you feel you’ve already spent too much money for the offers’ payout and it’s still hard to make decisions following the CPA vs Payout, then compare these parameters. Trust me: they’ll help you. Whereas CTR is useful to optimize banners and landing pages - since it measures their appeal - you can use CVR to see which offer performs best.
Just one more thing about the banners and LP: you need to analyse the best combination banner + LP. A banner with an overall bad performance may work very well with a specific LP.
Moreover - when deciding whether or not to cut - you have to not only consider the clicks but also the number of impressions and conversions of the banners.
Let me know more campaign details! That way, I can help you reach your goals!
Cheers!
Have to agree with
One way or another, its not really about how much clicks you need - when comparing just the offers - you need ENOUGH conversions for each of them in order to determine which one is the best.
The key is to set how many conversions is ENOUGH 
You can go for statistical significance for example, there are calculators that will help you to calculate which offer is the winner with X% probability. I personally don't rely on these tools too much, I simply want to see one offer take the lead by at least 25% - 30% to consider it a winner. So lets say I have 3 offers and the amount of conversions is like 6,9,13 - the 13 seems to be a winner, but 6 is a looser for sure, I would cut the one with 6 leads and run the other two for some more time. If the one with 13 keeps being better, I would consider it a winner and push just that one. Let me repeat, the key is to wait for ENOUGH, you can't do this when the number of conversions is for example 3,2,1 because all it takes is 1 random lead and all are almost even.
The numbers I used are just an example but Im sure you get the point I wanted to make.
Hey guys thanks for the input.
Yes the offers and LPs are in rotation (Im using Imobitrax for this) on a self serve network (traffic junky). I am in the US geo.
Yes I am only testing on 1 spot, I was under the impression that testing more is good?
Assuming you have a bigger budget and can ride things a little longer to get some statistical significance, do you still recommend just testing 5 banners?
So I should be testing 1 offer x 1 LP x 5 banners at a time?
Thank you in advance.
Originally Posted by jabong82
The style depends on what you prefer. A good number is 3-5 banners, 1-3 LPs and 1-3 offers - adjust based on your budget. As I said in my previous reply, before cutting anything, make sure you had ENOUGH conversions for every combination.
The only thing that dictates this is going to be statistical relevance.
https://help.optimizely.com/Analyze_..._to_run_a_test
Take a read of this article and you can make some decisions from there. The more variations and the closer performance the more data you need to make a decision.
Home > Paid Traffic Sources > Adult Traffic (NSFW)