Split testing is the practice of continually improving your website incrementally over time.
The way it works is you create two or more versions of the page you want to test, each one slightly different than the other. You “split” your traffic between these two versions and see which one performs better.
For example, if you’re selling a dog leash you might split test two different images: One of a dog on a leash, another of a pet owner who’s smiling.
After testing the images, you might find that surprisingly the pet owner smiling brings in more sales than the dog on a leash picture.
Instead of just creating a site and hoping that it sells, split testing allows you to increase your conversions over time.
Types of Split Testing: A/B Tests and Multi-Variable Tests
There are primarily two different types of tests you can conduct: A/B tests and multi-variable tests.
A/B tests are the tests we talked about above, where you test two different sites and see which performs better. It’s called A/B because there are two sites, one called A and the other called B.
The other type of test is called multi-variable testing. It’s a type of testing where you test many variables at once and use software to mathematically figure out the best combination of variables.
For example, you might test the headline, the layout, the picture, the price and the order button – All at the same time. The software will tell you which is the best combination of all the data.
Multi-variable testing is generally used only once you have a lot of traffic.
For our purposes, we’ll go over how to run A/B tests in this report.
What Should You Split Test?
There are a few things that affect conversions more than anything else. Start by testing these factors, then test less important factors later.
The price will often affect your conversions more than anything else. Should you charge $20, $40, $80 or $100? The only way to really know is to test it.
Be careful when testing prices. Generally you want to test it only for a limited time, or else you’ll be sending mixed messages to your market about what your price is.
If someone buys a product and later realizes someone else paid less, always refund the difference for them.
Next to the price, the headline is often the highest impact factor to test.
The headline’s main job is to capture the reader’s attention and get them to keep reading. It’s their first impression to your website.
If you spend 50% of your copywriting time on your headline, that’s time well spent. Split testing your headline can have very, very dramatic results.
The layout speaks to your users emotionally, non-verbally. By layout/design, I mean things like: font color, font size, width of the page, background color, etc.
Studies have shown that certain layouts can increase conversions quite drastically in certain markets, while reducing conversions in others.
For example, red headlines tend to work great in markets that aren’t heavily marketed to.
In markets like marketing, weight loss and dating tips however, people are so sick of seeing red headlines that a red headline actually reduces conversions. Instead, a deep blue headline which builds trust actually tends to convert best in these low-trust high-competition markets.
Each market is different. After you’ve tested your price and headline, start testing your layout to see what converts best.
Shopping Cart Dropout Rate
One thing marketers often forget to split test is the shopping cart’s drop out rate. In other words, once someone clicks “buy” what percentage of them actually completes the transaction?
Online, the drop off rate can be as high as 50% or more. With split testing, marketers have been able to cut down their shopping cart drop off rate by as much as half, effectively doubling their income.
The main things to split test in this area are:
1) Making sure the shopping cart page looks like the rest of the site, so a buyer doesn’t feel like they left your site when they hand over their credit card.
2) Safety. This is perhaps the #1 factor to lowering shopping cart dropoff. If you can help your customers feel safe giving you their credit card, they’re much more likely to be willing to pay.
3) Symbols like the “BBB Certified” symbol, “128 Bit Encryption” symbol and the “Hacker Safe” symbol have all been proven to increase conversion.
What is “Statistically Significant” Data?
One word that you’ll often hear in conjunction with split testing is “statistically significant data.” What does that mean?
It basically means you have enough data to know that if the test continued indefinitely, the results will be similar.
For example, if you flipped a coin 5 times and you got 4 heads and 1 tails, it would not be accurate to say that you’ll get tails 1/5 times by flipping a coin if you did it 1,000 times.
You just haven’t flipped the coin enough times to get statistically significant data.
On the other hand, if you run an A/B split test and find that out of 5,000 impressions one made 100 sales and another made 20, you have enough data to know that #1 will outperform #2 if you gave it a million impressions.
Any time you’re split testing, you want to make sure you have enough data that you’re getting statistically significant data. Meaning you have enough results to know that the data will stay constant when it’s scaled up.