Splittesting Ads and simultaneously splittesting landingpages (9)
Splittesting Ads and simultaneously splittesting landingpages
Hey guys,
do you split test different landingpages while you test different ads?
I do and it looks like that:
- 10 different target audiences, say f18-19, f20-21, f22-23 and so on
- every target audience has the same 100 ads
- 6 different landingpages for all ads and audiences
- tracking with facebook conversion pixel, splittesting with optimizely
All the campaign traffic is sent to the same URL regardless of the ad and the audience and then each visitor is randomly assigned a landingpage.
I see a mistake here:
- an AD might send the majority of its traffic to a bad landingpage. The ad may be good, but the CPL will still be high due to the low conversionrate on the lp
- there might be complex dependencies between audience, ad and landingpage. Some landingpages might only work with some ads or audiences
Is it correct that I need to split tests ads FIRST (100 ads, 1 audience, 1 landingpage) and then continue testing landingpages?
Cheers
Steve
I would always split test landing pages from the beginning.
Firstly, why would an ad get the majority of it's traffic sent to a bad lander? Presumably you would split traffic evenly over a decent number of clicks.
You are indeed right that there may be complex dependencies. If you have landers that convert similarly well it's likely that different ages/genders/demos match better with different landers. This is where you should use tracking to your advantage to uncover such relationships - how far you go is up to you and can depend on the capabilities of the tracking system. I call this deep path optimisation and it's one of the things I really like to nail - unfortunately most tracking systems are a pain in the ass when it comes to doing this, at least in the end when it comes to redirecting traffic.
My point... split-test landers and track like a beast so that no gems are left undiscovered.
Also, 100 ads is overkill - if you have time to make 100 you have time to make more than one lander. Start with something like 10-20 ads 2-3 landers. The ads are but one piece of the puzzle and focusing on them alone is shooting yourself in the foot (not so the case when direct-linking of course).
I would definitely recommend only cutting landing pages within a narrowly-defined demographic. So, for example, if you have 10 different audiences, I'd not recommend lumping all that traffic into one "pot" and cutting LPs based on overall performance: you're better to cut an LP for a single demographic if it's not performing for that demographic.
Thanks for the replies. I'll set it up like this:
domain.com/23-24/landingpage-1
domain.com/23-24/landingpage-2
domain.com/23-24/landingpage-3
domain.com/25-26/landingpage-1
domain.com/25-26/landingpage-2
domain.com/25-26/landingpage-3
domain.com/27-28/landingpage-1
domain.com/27-28/landingpage-2
domain.com/27-28/landingpage-3
I'll have a index.html in each directory (23-24,25-26,27-28) which contains an optimizely code which splits the traffic between the landingpages.
The facebook ads will be seperated on audience, so I'll create 100 ads for 23-24, the same 100 for 25-26, same 100 for 27-28. The ads only differ in the target URL.
Optimizely and Facebook have each two conversions goals: lead and sale, so it should be transparent which ads not only generate leads but also sales.
Yep, that's a sound basic approach.
I'd say that those age bands were probably a bit narrow - what a 24-year-old's life is like and what a 26-year-old's life is like aren't radically different. But obviously that may depend on the demographic and your knowledge of it!
Tip: If you just want to split traffic between landers you can use google content experiments which is free. Optimizely will start getting expensive once you are getting more hits.
For me, It depends on budget & traffic volume.
If you're spending $100/day and are split testing 100 ads, 3 demos and 3 landing pages it is going to take FOREVER for you to get any real data.
If you're spending $10,000/day you can learn real quick what works and what doesn't. *But i'd still never recommend testing 100 ads at at time. It's overkill.
What I normally do and recommend is test one thing at a time. It's much cheaper, and allows you to focus your spend on one campaign element at a time.
Example.
New Campaign Launch:
1. First, test age groups / gender using 1 ad and 1 landing page (preferably an ad and LP that have been "found" in the wild, because they have a shot at working)
2. Using the best age group, test ads for CTR (not 100, like 10-20 depending on budget)
3. Using the best age group and highest CTR ad, test 2-5 landing pages
4. Using the best age group, ad and landing page destroy some shit
It's a more methodical approach. Works for me. And might provide you with better results.
Oh, yeah, didn't see the bit about "100 ads".
I'd agree with Profitable - definitely test less than 100 ads as an initial test. 10-20 is solid.
Originally Posted by steve from munich
All 3 variables you mentioned have a MASSIVE role in conversion rate. 1 type of banner matched with the right landing page and audience will convert better then 99 others.
Also as said above, 100 is a few too many ads. Itll cost too much $ to properly test.
Home > General > Affiliate Marketing Forum