Multi-Variate Testing Versus Split Testing: You Can Always Take It Further
When you read the words split-testing, what do you think?
Do you think of a simple A/B test with two titles or pages to see which is better?
If you’re like me, then that’s what you think of.
However, there’s a whole ton more to split-testing that you’re probably not thinking of (certainly, I wasn’t.) In this article, I’m going to provide a framework which I got from this book which’ll teach you how you can really take your split-testing to a new extreme.
Split-Testing: Or, Multi-Variate Testing
I’ve written before that you only need to give a reader one reason to buy your product.
However, you don’t want to create one of those ugly HTML sales letters that go on forever. I’ve seen 30 page sales letters design to sell an e-book. That’s going to lose as many sales as it gains. The same goes for those video sales letters which are 40 minutes long and don’t let you buy until the end.
So, you need to test a ton of reasons to buy without listing them all in one letter.
The simplest way to do this is through multi-variate testing. E.g. you’ll have more than two different sales pages.
Strict A/B Testing works when you have a dichotomy. Does picture A or Picture B cause more click throughs? Does your target audience prefer moving away from pain or towards pleasure?
Those sorts of questions are easy to solve with a simple A/B test.
“Why are your customers buying this product?” is a whole different story. You’re not going to get a good answer for that from an A/B test.
Let’s take skin cream as an example.
Think of the major reasons why a person would buy a cosmetic skin cream.
Look better?
Feel better?
Look younger?
Be less self-conscious?
Have less sunburn?
Those could all be valid reasons. Often, you can tell based on various things which is going to work best. However, sometimes you’ll be guessing.
Even if you’re not guessing, the more you drill down and get into specifics, the more you will be guessing.
For instance, “looking younger” is probably what a 40 year old woman wants from her skin cream. But does she want less wrinkles? Does she want the crow’s feet to disappear? Maybe she’s worried her cheeks aren’t as full as they were when she was younger?
All of these come under, “look younger” but one of them is probably going to be more successful as a headline than the others.
So, you decide to do multi-variate testing.
Multi-Variate Testing: How Much Is Enough?
You might read the above and realise that you could test every possible combination and get yourself into a world of complexity. Some of you might be into that, but I’d bet most of you want to find what works best and stick with that.
I certainly do.
When it comes to multi-variate testing, there’s one thing to take into consideration: Everything needs to flow.
So, if you create an advert saying, “Learn How To Talk To Girls” then the sales page you send it to is going to have to be about talking to girls. It has to be on the same subject, pitched at the same level… there’s no point in sending a “Learn how to talk to girls” clicker to “master seducing multiple women at any time” advert. There’s also no point in sending a “master seduction” ad reader to a “How to talk to girls” product.
The way to do this then is to create a big idea for each sales page, and then miniature angles that spring from the big idea.
So a “Look Younger With This Skin Cream” sales page could have the wrinkle ad, the full cheeks ad and the crows’ feet ad. You wouldn’t have a “How to look younger” sales page with a “Look better in a bikini” ad, because you’re selling to a different idea then.
Similarly, you could align a “Scared of talking to girls?” ad with a dating service sales page, but you wouldn’t use a “Scared of talking to girls?” ad to a casual sex site landing page.
Then What?
That’s the first key to working out how much material you need.
The second is coming up with those big ideas – this is what your product is called/selling overall.
The third is creating the niche needs that come from the big idea – those are what your chapters/sections/aspects of your product sell.
The fourth is to then work through this all and create a pyramid. This is because you don’t want to write 10 sales pages – one for each Google Ad or whatever – when you could write a single sales page and write the ten google ads so that they are relevant to the single landing page.
Essentially, you want to create a pyramid structure where you can write a ton of short ads, then less platform specific landing pages, then even less product landing pages.
This way, you can find out pretty quickly which ads work and which don’t. You’ll also find out which big ideas work and which don’t.
It’s only once you’ve tested these things that you can work out how to actually optimise your sales letter. After all, without going through this process, you won’t know why people are buying or not buying your stuff.
Final Thoughts
Multi-variate testing is pretty straightforward. I’m not sure I’ve explained it in a straight-forward way though. (If I’m rambling, leave a comment. I’m working this all out as I go along.)
When it comes to building a business, data is key. In fact, if you go for venture funding or anything like that, you’ll be expected to do tons of market research and boring questionnaires and stuff.
You could do that, or you could write a few sales pages with “Business Coming Soon” signs at the end of the funnel. That’ll give you actual real world data that’s more relevant than any market research.
P.S. In the book I mentioned above, the author recommends 5 big ideas (known as angles in affiliate marketing) 5 landing pages (one for each angle) and 5 ads for each landing page.
I find that that’s a bit prescriptive, but if you’ve got an idea and want to implement it, you could probably do worse than following that pyramid structure.
You can check my review for the book here. You can also wait and see what I come up with when I try this out myself in the next couple of weeks.