Split testing has grown in popularity over the last few years as a quick and easy way to increase your conversion rates. It is a method of testing multiple versions of a website whether it be a homepage or a landing page to see which one converts the best. During the test, any traffic to the site is split between the different versions that are being tested and the performance of each version of the site is tracked for comparison. Site metrics that can be compared may be clicks, opt-in form completions or online purchases. The information gained from this sort of testing can be extremely valuable for those who wish to increase their conversion rates.
However, not every split test is a good split test. You will want to make sure that you use a good strategy to avoid testing useless and irrelevant factors. There are many things you should consider when planning a split test to get the most out of the process.
Test One Variable at a Time
The first thing you should consider is the number of variables you are testing. In practically every case, you should only be testing one variable at a time, therefore only making one change to your site. It is tempting to test a host of different variables at the same time in order to “make the most” of your split testing. However, the value you gain from that sort of split test is less than what you gain from a more systematic testing of one variable at a time. Testing only one variable allows you to pinpoint the exact reason why one variation of your site was more effective than the other.
If you change many different variables, you can test to see which version of your site converts best, but you are missing out on some important information about exactly why the better-converting version had a bigger impact.
When you know which variable made a difference, you have some valuable information about your customers and what makes them more or less likely to convert. This will be very beneficial when you undertake other marketing campaigns in the future because you can apply those specific changes that you know are effective and avoid less effective measures.
The importance of limiting the variables that you are testing extends to the color choice, web copy and images used on your site as well. They should be exactly identical to each other except for the single variable that you are testing so you can ensure that a change in color or something else did not overshadow another variable you were actually trying to test.
Duplicate your Page to Save Time
In order to run the split test, you will need two similar websites with some, likely simple variations. Once you have finished building your landing page, you will need to create a variation of the page to run the test against. Rather than putting a great deal of time into creating a new page from scratch, you can simply duplicate the original and then change the elements that you wish to test. This seems obvious, but it is important to remember.
By duplicating your page first, you are saving yourself a great deal of time but also ensuring that there are not mistakes made that result in the variation page having other changes made to it and therefore leading to testing elements that you weren’t planning on testing.
Choose the Appropriate Period of Time
When running a split test, it is important to carefully consider timing. You will need to decide both how long you wish to run the test for as well as when exactly to schedule the test. Both decisions can have an impact on your results.
In most cases, it is acceptable to run the test until the page reaches a level that is considered validated. This is typically when the likelihood of outperforming the original page or the variation page is more than 95%, or at least near there. There are some instances where you should consider running the test even longer than that. First of all, you want to make sure that your test runs for an entire week at the very least. If not, you are adding another variable of weekend performance versus that of weekdays to your test. Running the test long enough to at least get a full week’s worth of data is important for the accuracy of your results. Along the same line of thought, if your test has run for nearly two weeks, but has not yet reached the two-week point, it is advisable to let it run until you get a full two weeks-worth of data. Just like above, this will keep the week-day and week-end ratio the same as a normal week. This is important because the things that people are motivated by can vary greatly depending on whether it is the weekend or a normal weekday.
You will also want to make sure that you choose a good time to run your test. For instance, things like shopping seasons, holidays and promotions all affect the results of your test. For instance, if you run a split test during a huge black-Friday sale, the information you will gain is not representative of how your customers will perform during a normal time-period, but rather will be representative of how they will perform during a huge retail-shopping event. It isn’t always a bad idea to perform the split test during these special time of year, but it is important to realize how it will impact your results, and if you are measuring what you want to measure in your test.
Split testing is an incredibly valuable tool and when done right can have a huge impact on your conversion rate. By following the strategies outlined above, you will be better equipped to run effective tests that improve your marketing strategies in the long run.