Today’s topic is how to split test or do an A/B test with your marketing.
So first off, let’s talk about what are some of the things that you can insure doing a split test for. A split test is simply taking a set of creative pitting it against something else. And when you’re doing an A B split test, the goal is to start to look at which of the two is performing better. Now when I say creative, we’re going to get a little bit more finite as the episode progresses. But that could mean radio episodes or radio commercials.
You want to split test. It could mean landing pages, it could mean two different homepages on your website to see which one works better than another. It could be testing direct mail, flyers, door hangers, literally anything that you’re doing from a marketing standpoint can be split tested.
Now the most common example of a split test is a landing page. So you’ll leverage paid traffic like Facebook ads or Google ads and you’ll drive for a certain portion of people to one landing page and a certain portion of people to the other. Now there’s a few things that you have to make sure that you execute for this split test to actually be valid.
The first thing and the most important thing is you can only test one variable per split test. So for example, if you’re going to test two different landing pages, you might want to split test the headline at the top versus a different headline on the other landing page. But what you can’t do, if you want an accurate split test is one landing page you have a certain headline, the other one has a different headline. But on the second page, it’s a different color. It’s a shorter copy. They’re more testimonials.
You have to make sure that it’s apples to apples and you’re looking for one thing to another. Now, that doesn’t mean that you can’t have multiple split tests, but you have to focus on one variable that you’re changing. And normally, when you’re doing a split test, you’re going to aim to really split test, something that you’ve put a hypothesis towards.
For example, if I change the color or the background on one of these versus white, I just have a feeling it’s going to convert better. So that could be one of your split tests.
Now, the other critical factor is that for the split test to be valid, you have to make sure that there’s an even amount of people that are seeing each of the different pages and that’s again, we’re assuming you’re testing a landing page. If 400 people look at one page and 100 look at the other and you say well based on these metrics, it looks like this one worked better. Well, that’s still not an even split test.
So step one is you’ve got to pick what variable you’re going to test. And normally, that’s a headline. That’s a different call to action. So maybe everything is exactly the same. Except for one landing page, you’re saying the call to action is to call the other landing page. It’s to text in, for example. Then what you’re doing is you’re split testing to more people call or the more people text. And what you’re doing is as you’re executing these different split tests, you continue to test more and more variables, again, one variable at a time.
But if you start with one page, and then you design the other one, and let’s say you start with the headline, okay, great. You run that for a week or two, you declare a winner. Then you start working off of that landing page. Now what do you want to split test next? Is it the color, images, calls to action? You pick up the next variable.
Let’s say you’re going to test that whole calling or texting. So the first landing page, it’s going to have call, the next one will have text, you’re testing now the same headline because again, you just picked a winner from your last test. Then you pick a winner with that, then you can move to another test.
So you’re always innovating and testing and the whole goal is to consistently one up yourself. You always want to continue to perform better. So there’s really never a time when you would not be conducting an A B, split test. Because you’re always able to test something else that would have been one word, one different color. Things like changing the call to action button from red to green could make the difference of a 10% spike in your sales.
All the little things actually matter substantially. If you wanted to split test something physical in the mail, what you would do is the same kind of concept. If you were doing direct mail, and let’s say you had a list of 1000, you would send 500 people, one of the postcards, 500 people the other. And again, you’d be testing only one different variable. You can’t send two completely different postcards. That’s not an accurate split test if you would test one different variable.
If you wanted to test something with your door hangers to see. Do people want to visit a website? Or do they want to call and schedule a quote? Maybe 100 houses you put one of them, the other hundred houses you put another one? When was the last time you did something like this? The chances are pretty high that the answer is never. This is normally not something that local businesses are thinking about. But this is where you’re leaving money on the table because there are always ways to improve.
You don’t need big fancy marketing agencies to do all this creative for you. You could sit in a room brainstorm and say, let’s test this. Let’s test that. Let’s test a different color, a different logo, a different testimonial, maybe it’s testing a family testimonial versus a happy couple testimonial. I mean, there’s lots of different things that you can test. And then you continue to leverage that data again to up yourself and get better and better and better.
I’d also encourage you to do this, if you’re running radio ads. If you’re doing billboards, you can’t just run the same thing consistently and expect that it’s going to always improve.
Now, that’s also not to say that what you’re going to test is going to improve it. You may run one landing page, you think you’ve got the best headline. You run another one that’s got a modified headline you think is even better and it could perform 50% worse.
When you go back and maybe you try a different headline, and you test that, maybe there’s a different variable. So there’s always lots of different variables that you want to look at. But when you start to do this, I promise you, you’re going to start to see some pretty big differences between one offer, another offer, one headline, a different headline, one type of testimonial, another one, a certain color, for example. But again, test one variable, and you also have to keep all the other factors essentially the same.
So you can’t effectively do it if you’re testing a radio ad, you can’t be testing a radio ad that’s running at drive time, like 7am versus one at 10pm. And the one at 10pm has a different offer. And that’s what you were testing. It just doesn’t work like that because the 6am audience is vastly different than 10pm. The amount of listeners is vastly different. You’re not comparing apples to apples, you’d want to run one offer at drive time for three to five days, for example, and then you’d want to run a different offer only modifying the offer or the call to action for another three to five days the following week and compare the metrics.
But also, again, make sure that you’re comparing apples to apples. There’s no holidays involved and things like that to ensure that your ad is equal.