It’s hard to say exactly when the first email split test was run, but it’s been around a fair while. AI is the new kid on the block. Is new better than old in this case? Or does experience beat innovation? .

Split-testing as a concept is much older than email. It simply means offering too slightly different options to a sample of your audience and seeing which one gets better engagement. For example, one email might have all its text left-aligned and the other centre-aligned. Studies of psychology and eye-tracking have shown that even banal changes like this can make a noticeable impact.

AI, or Artificial Intelligence, is a recent entrant to the marketing universe. We’ve heard a great deal about how it is going to up-end all marketing functions. What we’re talking about here is a tool which can scan a huge volume of previous emails and identify particular elements which impact the engagement rate. This could be the length of the email, the Sender Alias, or the picture-to-text ratio.

Both tools offer a variety of pros and cons, so let’s take a look at each in more detail.

Why split testing is beneficial for marketers

Even if you aren’t running split tests on everything you send out, you should know what split tests are and how they work. Create two versions of your email, with one difference between them. Send each one to a test segment of your audience. Whichever one performs best, then gets sent to the whole audience.

A major advantage of split tests is that they are working with live data. AI uses previous engagement metrics to calculate what the recipient will do. If something big has changed, like a corporate takeover, then AI will have no way to take this into account.

Along the same lines, what about new contacts? You’ve come back from a particularly successful expo with armfuls of new leads. You add them into a followup campaign, but AI can’t tell you anything about them, as it has never seen them before. A split test is a far more effective way to start learning the behaviour of new contacts, wherever you have got them from.

Split testing is a good way of measuring specific changes. AI can tell you whether more or fewer images would be good for your email. It can’t tell you whether having a person or a cityscape as your banner image will perform better. A split test will show you very clearly.

The one main downside is that you can only change one variable at a time in order to get clear results. If your two versions have multiple differences, it gets hard to tell what level of impact each one is having by itself. This means that split testing is time-consuming, as each change requires a new test.

What happens if you get a draw? If both versions of your email receive the same levels of engagement, you’re back where you started. Split tests intended to resolve creative disagreements within your team can often fall foul of this, leaving you with effort spent but no tangible gain.

Ways in which AI is helpful for your marketing

Artificial Intelligence in this context is a piece of software which can analyse your email for you, taking previous engagement rates as its reference material. It can look at the number of words in an email, how long the sentences are, and how many images an email contains. It then compares this to everything else you have sent out, and makes recommendations. While AI has a way to go, there are already several areas where it represents a great leap forward.

The greatest advantage of AI over split testing is that one test can flag multiple improvements. As opposed to painstakingly split testing one variable at a time. Our GatorAI tool will also rank the improvements, so you can see how much difference each change makes. If three extra images would only improve engagement by 0.001%, it’s not really worth it. If shortening a couple of sentences would make a 9% difference, what are you waiting for?!

All AI testing takes place within your Gator Account. This means that you don’t have to bombard your audience with multiple emails as you split test different functions. If you space your split tests out, you can end up spending all your time on one campaign, neglecting your others.

The real power of AI lies in the number of results it can draw on. Our AI has been fed over 2 years’ worth of data, with millions of emails to all sorts of different audiences. Going through all of these manually to learn from split tests would be painful! The other benefit of analysing this volume of data is that AI may flag a change you don’t even remember making. Either because it was done by a predecessor, or just so long ago you’ve forgotten.

AI’s problem is that it relies on historical data. Let’s say you have just launched a new product or service. AI will be unable to account for this, especially if you are using terms that have never appeared in your sends before. While it can still tell you if your sentences are the right length, its ability to assess the language you use will be severely limited.

Striking the perfect balance

Given the levels of both interest and investment in AI, it will soon catch up with the areas where it is currently lacking. However, at present, it is best to use both tools in tandem. Use AI to make sure your email has the right number of images, then set split tests to see what image your audience react to best.

It’s dead easy to set up and run Split Tests and AI tests within our GatorMail platform. Download a copy of the brochure to see how we’ve built everything you need to test and improve your emails until they are perfect.