Primer on split testing and cost per order

Mrsmiley in last blog asks what split testing is, as well as important question about measuring results. You need to know this stuff as a web developer/designer if your market involves helping people convert visitors to customers.

A split test is a way to statistically test two different offers and see which is better. Here’s how it works:

1. Create two different URLs for the same product (e.g. www.URL.com/1 and www.URL.com/2).

2. Have one difference between version 1 and 2 (in practice, you might test a few small things at once). Test big things, like price or payment terms or design.

3. In your advertising, send people to both sites.

4. See which ones gets better conversion.

Note that direct marketers and folks with huge volume routinely create much more complicated matrices, in which they test lots of different offers on lots of different URLs.

For more info, subscribe to free newsletters at www.marketingsherpa.com.

Now, onto conversion and cost per order. Mrsmiley asks why I care if one site costs $10 to get a customer and another costs $100. Aren’t both sites different? Well, yes and no.

In my case, my sites cost about the same — $45 per quarter. And let’s say that most subscribers stick around for 3 quarters, generating $45 X 3, or $135 in revenue for me. So I can pay $45 X 3, or $135 for a customer to break even (assuming there are no variable costs once the software is developed).. If it costs me $10 per customer for one site (e.g. www.attractnewclients.com), and $100 per customer for a second category (e.g. www.commonsenseparentingonline.com), I am going to market both since both are profitable over the life of a customer. But I am going to invest heavily in the $10 per customer site, since that customer is profitable right up front, on first payment of $45.

BTW, the above numbers are not real, just illustrative.

Too many web designers are focused on SEO and getting people to a site. That’s easy compared to converting visitors to a sale. It is MUCH more important to know how to convert customers.

I hope this helps!

Win an Annual Membership to Learnable,

SitePoint's Learning Platform

  • http://www.realityedge.com.au mrsmiley

    Speedy service, I only posted on that topic about an hour or so ago. I misinterpreted the $10/$100 part as your charge to the customer, not cost to you to get the customer. In that light, it obviously makes a lot more sense.

    Oddly enough I’ve been subscribed to marketing sherpa and never heard the term split testing, although read a bit on the technique.

    Thanks for the explanations. I’m about to launch a new product website and will look into this split testing technique a lot closer.

    When you say send people to both urls, are you talking about within the same piece of advertising, or via different pieces? eg. say 500 flyers with one url, another 500 with another url, etc.

  • aneitlich

    Mrsmiley:

    Everything in a split test should be identical (except the thing you are testing). So if you are testing price or an offer, then you want customers from all of your ads to go 50% to one URL and 50% to the other.

    If you are testing 2 forms of advertising, then you should create a system so that you can track their behavior by ad. The way I do this is by creating separate URLs with the same content (called a campaign). Ad1 takes customer to URL1 and Ad2 takes customer to URL2. Then I can track behavior more simply.

  • ElemZero

    When you say send people to both urls, are you talking about within the same piece of advertising, or via different pieces? eg. say 500 flyers with one url, another 500 with another url, etc.

    You don’t even need to use different URL’s… There is software available that will show version A or version B of a page at random, track and compare conversions, etc. There are a few packages like SiteSpect but also try a web search for ‘split testing software’.

    Good luck!

  • http://www.redflystudios.com cianuro

    I use split run test software, and agree.You do not need to use different URLS and monitoring can be graphed easily too.