Best practices for putting angles in banners (5)
Hey guys,
So I was thinking about the best procedures to create banners. I read it was suggested that we start with 2-3 angles and from those we create 5-10 unique banners (total 10-30 banners).
Shouldn't we first test as many angles as possible? If we were to use a generic banner template and test 10 angles instead, we would be able to filter out which ones are best in terms of conversion. Isn't what affects conversion the most the headline/angle? From there we could then pick the top 2-3 angles to create varying banners. Could someone confirm me whether my train of thought is right or wrong, I'd appreciate. 
That's sensible thinking, but ends up with a few problems, which is why we recommend working the way we do.
If you "test" an angle by running a single banner for that angle, all you're actually testing is that single banner. For some angles, some banners will work better than others. Even if you use a generic banner design, you've still got variance in the image and text you use - plus, some angles will work better with some banner designs.
By testing a variety of banners for each angle, we make sure that when we test our angles, the results we get are more representative - a single accidentally terrible banner won't sink an entire angle.
You raise an interesting point: "Even if you use a generic banner design, you've still got variance in the image and text you use"
This is especially true in Adult, where the image is SO important. I find that there are so many variables that it can be difficult to hone in on what's actually driving success / contributing to failure.
I tend to favour running 3 - 5 ads per angle with the same headline, CTA, text and design, but with different images. But a part of me still thinks that isn't enough, as there's still a chance that all the images were rubbish. Also, when you test the next angle, do you use 5 different images, or the same ones as last time? So many variables...
Any advice on how to approach this?
@massmarketer - it's a tough question. The way I look at it is that it's less about making sure you'll never be fooled by poor images, and more about using enough that on balance, fails are probably related to angle or offer, not images. Hope that helps!
Awesome discussion! What I personally like to do is this...
1st Round of Testing
I would test say 5-10 angles, 3-4 banners/ea, each banner with a different headline. Then I would find some generic images (i.e. not angle-specific) that I believe will work well and use each image across several different angles.
This way I can sort of see a trend of which angles and images are working. And then I would take all ads that show enough promise (this part is subjective), and create combos from their headlines + images.
So for example I have:
5 angles
3 headlines per angle - HL1a, HL1b, HL1c, HL2a, HL2b, HL2c, HL3a, HL3b, HL3c, HL4a, HL4b, HL4c, HL5a, HL5b, HL5c
Found 4 images - IM1, IM2, IM3, IM4
I would create the following banners:
B1 = HL1a+IM1
B2 = HL1b+IM2
B3 = HL1c+IM3
B4 = HL2a+IM4
B5 = HL2b+IM1
B6 = HL2c+IM2
B7 = HL3a+IM3
B8 = HL3b+IM4
B9 = HL3c+IM1
B10 = HL4a+IM2
B11 = HL4b+IM3
B12 = HL4c+IM4
B13 = HL5a+IM1
B14 = HL5b+IM2
B15 = HL5c+IM3
Then I would look at the overall/combined stats for B1,B2,B3 to evaluate how good angle 1 is, B4,B5,B6 to evaluate how good angle 2 is and so on...
And I would look at overall/combined stats for B1,B5,B9,B13 to evaluate how good IM1 is and so on...
2nd Round of Testing
Then as a second round of testing I may do the following:
-Take promising angles and write more headlines to test them.
-Take promising banners and make different combos of their headlines and images. For example if promising banners are B3, B8, and B11, I would mix and match their headlines and images as follows...
B16= HL1c+IM4
B17 = HL3b+IM3
B18 = HL4b+IM4
(Note: B3, B8 and B11 have 3 unique headlines and 2 unique images used across them. 3 x 2 = 6 combos, minus the combos used in the original B3/B8/B11 banners. )
This way you become more and more certain which angles, headlines and images are better than others.
3rd Round of Testing
As a third round of testing I may do this:
-Take promising headlines and test them with a bunch of new images - this time I can include images that are more relevant to the angle.
-Come up with completely new angles and headlines and combine them with the best generic images I've found so far.
Hope that helps someone! It may not be the best strategy but if you don't have one at all it wouldn't be a bad place to start. Then you can improve on this strategy as you gain experience.
Amy
Home > >