Help me with my testing, How many LPs should i test at once, 18? (11)
Help me with my testing, How many LPs should i test at once, 18?
Hi guys i am struggling with some testing.Here is the thing..
I have 18 different landing pages.
8 of those are unique angles.
The other 10 are one proven angle which i am selling with slight differences in the wording and the design layout.
I figured out that first i should test the 8 unique angles. But i am not sure that i will have significant data if i test all the 8 lps at once.
I am using interstatial traffic, something like pop ups. Lets say that with my Baseline Landing page i am geting around 100k views DAILY, with 1.06% LP-CTR and 1055 clicks to the offer, 43 conversions at 4.2% Conversion rate, the revenue is $90.00 and the profit is 20$ the offer payout is $2
100000views / 8lps = 12500 views per LP and if the ctr is atleast 1.06% i will be geting 125clicks to the offer per lp. So am i gonna be able to get statisticly significant results and determine a winning lp with only 125 clicks to the offer and maybe 5 conversions ? I think no...
So how many lps should i test at once with my traffic volume so i can find my winning lp as fast as possible.
If i start testing 8 lps at once will i be able to find atleast some clear winners and some clear losers ?
How to determine how many lps to test at once based traffic volume ?
Damn i am confused..
Also what about that calculator is it gonna work http://www.conversionvoodoo.com/a_b_...calculator.php or there is big chance of error like with the traditional statistical significance calculators?
I'm also a newbie so I could be wrong.
you have 8 landers with unique angles, and then 10 landers with a PROVEN angle with different variations of wording - from what i see these are 2 different groups of testing. test the 8 unique to see which angle works best, and in a seperate test do the 10 landers with a proven angle.
Yeah i figured that. Currently i am wondering can i test the 8 lps all at once and easily discover statisticly significant data and find clear winning lps and clear losing lps?
So i can cut the losers and continue the test with the winners only.
But what if there aren't very clear winners and my data always shows that i need to test more, i will be losing lots of monies, hmmmmmm
Maybe i am making things too hard for myself by trying to over complicate something that can be done in more simpler way.
Currently i want to do A/B/C/D/E/F/G/H test (testing 8 lps all at once) but i am not sure it will be cost efficient, if my data doesnt shows very clear winners fast. Hmmm
If all lps perform in similar way i will need tons and tons of traffic to find a clear winner (statisticly significant). Right ?
Maybe i should only do A/B tests. Where i test Baseline lp vs. lp1 = find winner. Test the winner against lp2 = find winner. Test the winner against lp3 etc etc.I am thinking that this will lead to more clear results.
Give me some input guys..What will be the best way to test in my situation ?
You mentioned you have 8 landers with different angles. Before the landers, have you tried testing those angles on the actual ad to see if people would click it - see if that angle actually works?
If no ones biting on the ad angle, then wouldn't it be inefficient to test the landing page angle? At least this way, you can narrow down the amount of landers you have to test.
Well i am not using ads. I am getting my visitors straight to my LP. Good suggestion though. 
I would split test 3 proven angles, and 3 new angles. Hold onto the data, then split again with new pages. 18 at once is a lot, and even 18 over a period of time is a lot. Just make sure your testing with a lot of money so you can acquire a lot of data. Cutting those 18 pages too short will give you technically no results at all.
yeah that's kind of a lot. Have you looked into Landing Page Genius? Maybe you might want to consider looking into taguchi style landing page testing.
I'd strongly recommend using the calculator I wrote about last week for this sort of test: it'll avoid compounded significance errors (as you mention earlier) and give you better results faster.
As Stackman says, definitely don't test all of your LPs at once: split them into batches and test them. Apart from anything else, the first lander you test might be a knock-it-out-of-the-park winner, and you don't want to have to buy data for 18 LPs just to discover that.
Multivariate testing, as Jasper mentions, is also a great idea for PPV-style campaigns (which is what you're running, I think?). Definitely look into that.
However, if you're testing a bunch of wildly different angles it won't help you initially (at least not without some wicked complicated math behind it). It'd be great to optimise further once you know what angles work, though.
@Jasper yeah i have LP genius but multivariate testing is for lp elements only right ? When i find a winning lp i will use LP genius to tweak it further, thanks for the suggestion
@Cauerman yes i am using the bayes calculator that you posted about, and also extensively reading and learning your latest threads about testing.Thanks for those
@Stackman yeah i am doing something similar now A/B/C/D test completely different angles
Personally I would test 5 at a time, replace the worst 3 with 3 new ones until you are done.
Then decide from that data.
Home > Technical & Creative Skills > Tracking Campaigns