Phenomenal - Split Testing Sales Letters

[USEO]

Hi Warriors,

This has blown me away. So much so I wanted to share it.

We're all busy trying to make as much money as possible right?

I'm sure to some other pro's on here this is a no brainer and it has always
been something I've been meaning to test on a few sales letters.

For noobs, you really must think about this...

I've just been split testing a sales page using a cheap bit of script (not
selling anything here, just look on Google, there are tons available)...

The reults for like-for-like traffic have been staggering.

Set up one of these scripts and play off headline 1 against headline 2...

Establish the better one, bin the poorer one and add a new one...

Repeat...

You can do the same with your deck-copy, first paragraph, price - pretty
much whatever you want...

But DO IT!

For like for-like-traffic on the previous week sales were almost double.

Like I say - split-testing is a no brainer once you use it, but I'm just
wondering how many noobs who put bucketloads of effort into selling -
actually use split testing...

Hope that's good food for thought.

Regards

Craig K.

[Brad Gosse]

Agreed a good split testing is worth a fortune in conversions

You need a ton of traffic to do this properly.

[MikeHumphreys]

Originally Posted by Brad Gosse

Agreed a good split testing is worth a fortune in conversions

I agree 100%. If you are not testing then you are leaving a bunch of money on the table instead of putting it into your bank account.

You need a ton of traffic to do this properly.

Not necessarily. It depends on what you are testing and what types of conversion rates you're seeing. If you are using any of the paid testing tools, they have a confidence level stat built-in that tell you how likely the current testing data will stay the same.

The general rule of thumb that I recommend is 100 actions. So if you're testing a squeeze page, we're talking 100 opt-ins. A squeeze page converting at 50% needs 200 unique visitors max to get you the 100 minimum actions for reliable testing results.

Hope that helps,

Mike

[USEO]

Quote: You need a ton of traffic to do this properly.

I don't disagree because more data as rule ='s more reliable results but...

You have to start somewhere.

What I'm saying is that my recent experience of split testing has shown
how important it really is to maximising income.

If you are getting 10 vistors a day and all 10 buy from headline 1 while
headline 2 delivers zilch - it's a no brainer in my book even from 10.

It's like tossing a coin...

Even small traffic will give you results.

Thanks for your input.

I'm suprised more people didn't comment.

Maybe it's just that I was only me not split testing before :-)

CK.

[damongreene]

AMEN! Too often business owners let their pride and ego stand in the way of good conversions. Just remember. The site if for your VISITORS and CUSTOMERS not you!

What better way is it to find out what they want then asking them... and paying attention to what they respond to. Im also a HUGE advocate of surveys! Stop guessing and KEEP testing

[d101]

... http://google.com/websiteoptimizer

[Catalin Ionescu]

Originally Posted by USEO

If you are getting 10 vistors a day and all 10 buy from headline 1 while
headline 2 delivers zilch - it's a no brainer in my book even from 10.

It's like tossing a coin...

It's precisely like tossing a coin. You can't reliably say it is biased even if you get a streak of a few tails in a row, can you?

That is why even if you see all your 10 visitors from a day buy from headline 1 you shouldn't jump to the conclusion it converts better than the second one. It may have been a fluke, or some other circumstance like perhaps an affiliate promoting a better message and all of his subscribers end up on version 1...

You should wait a bit more until you can assign meaningful conclusions to the numbers, 10 actions is very very little to say anything about it.

I won't get into the maths behind, but as a rule of thumb you should wait until you get 100 actions or more for a split test. For multivariate split testing it gets more complicated, but any software worth his salt will show you the confidence levels you can place on the numbers.

Sincerely,
Catalin Ionescu