YOLO! split testing techniques? (9)
HEY guys. I been apart of the STM forum for few months now and Im pretty sure I got setting up and launching campaigns using
Just pick the variable you want to test and make exact duplicates of ads/campaigns with that one variable being the only difference.
When there are lots of things to split-test, focus on doing the variables with a larger impact on audience demographics or traffic intent first (e.g. split-test wifi vs carrier, or male vs female, or ages 18-21 vs 22-25, rather than split-testing red banner text vs blue banner text).
Originally Posted by zeno
The problem is that the extent of impact of a variable itself can vary with traffic source/vertical.
In general, I would suggest looking at it like this:
Highest impact: the angle (involves many creative aspects...), the traffic source, the offer (and where you run it, e.g. networks).
Medium impact: characteristic targeting variations at traffic source (e.g. ISP, device type), traffic source placements/pubs, landing page designs, banner designs.
Low to moderate impact: absolute bids, banner/lander minor variations e.g. colours/non-critical text, time of day and day of week, etc.
It's actually pretty hard to class these as depending on the particular campaign, time of day for example might move into the high impact category.
Ideally your tracking would take care of a lot of these in terms of identifying exactly what has a large impact.
Then, you can focus instead on creative split-testing - angle in particular, and targeting parameters that you know have a large effect based on what you have learned historically for that traffic source.
Not sure that's particularly helpful but should make it clear why buying data and getting pure experience is a critical part of the whole process.
hmmm... okay I think I understand the basic concept of split testing. I think I am going to start by split testing offers first. When doing so would I basically have the same ad creatives, demo etc. and the only different variable would be the offer? Would I direct link or use a lander? Also would I just let it run for about 2000 impressions until I pause and change variables to which ever gets the most clicks or wait for more data?
Hey zeno could you possibly share some creative aspects of the angles? Ive read many blogs (dr.ngo etc) and forum posts but I would just like to see what your 2 cents because seems like everyone focuses on different aspects to angles. Thanks tons man.
For split testing offers just rotate them evenly using your tracking system. You do not change your creatives as part of that. Whether using a lander or direct-linking is irrelevant, the choice to split-test offers is completely independent of the choice to use a landing page or not - as is virtually all split-testing (you isolate and split-test variables separately from other variables).
I'm not sure what you're referring to with 2000 impressions but that's a metric for creative views - which has nothing to do with split-testing offers. If your creative is seen by a user, so what? The data that matters for offer split-testing is how many clicks you send to the offer and how many convert. If your ad creatives aren't getting enough clicks to the offer for this split-testing to be effective, then you just need to focus more on the front-end first than the back-end.
When you get to the point that there are statistically significant difference between offer performance (earnings per click!), that's when you might remove the weaker from rotation.
As for creative aspects of angles an angle is how you frame your offer and present it to your audience. So it obviously involves the banner, it's imagery and it's ad copy, as well as virtually every marketing component you expose them to on their way to conversion, i.e. the funnel.
The main creative components are the banner and lander imagery and ad copy, and this is a very, very broad categorisation at that. Imagine asking someone to 'explain the technical aspects of a car'.
Thanks Zeno. That's what I was pretty much referring too the part where you explained that what if I don't get enough clicks to my offer page and no one converts? Because I'm just starting out on POF and still haven't seen to get a conversion yet. Does this simply mean my ad copy and image for ad is simply not good enough? In this case wouldn't it be a better option to just direct link my ads to the offer page and not even use a lander at this point to find what offer converts better?
I think you should use a landing page from the start. PoF is competitive and not have a lander simply won't cut it in the majority of scenarios.
If you don't get many ad clicks, try to improve your CTRs by testing ad variations and angles.
If your lander CTR suck, test multiple types and designs.
If your offer conversion rate sucks, split test multiple offers and your entire campaign angle.
If you're getting some clicks to the offer (20+) and they aren't converting then it could be many things, but most often it's your angle or the offer itself.
To put it more frankly, your campaign isn't good enough at that point, but there are multiple points where you can improve it.
Whether you try to - and which parts you focus on - really depends on your data as well as taking a step back to assess overall campaign spend and profitability, and the probability that the current iteration of the campaign will get you anywhere.
Home > The Newbie Zone > Questions and Answers