How To Get The Most From A/B Testing

How To Get The Most From A/B Testing

A/B testing is the practice of changing two variants in online advertising in an attempt to maximise conversions and gain more sales. It’s something companies do regularly. There’s a right way and a wrong way to perform an A/B test, though. If you aren’t maximising your conversions, you’re doing it wrong.

Picture: Marc Levin
A/B testing is ultimately about using a landing page to maximise conversions. It’s not about bringing as much traffic in as possible. The success of advertising is based on sales, not on browsers.

What Are You Doing For Returning Customers?

Marketing executives are always talking about how they can bring new customers in. What they almost never talk about is what you’re doing for returning customers. How are you going to keep your content fresh for them?

Returning customers are your most important customers. Those are the people you can rely on to purchase your new releases. New customers might buy one thing and never return. Your A/B testing should take into account what it’s going to do for people who have dealt with you before.

Focus On One Variable

Although A/B testing focuses on two variables, you should only ever change one variable per test. Using a new call to action and a new landing page at the same time might offer results, but it won’t show you what’s responsible for it. Both factors might be responsible for it, but you can’t know this.

This form of testing is about combinations and getting the right combination. You can only manage the value of an individual variable if it’s used on its own with something you already know the value of. This existing variable will be your base value you can measure everything else against.

Have A Plan

Other variables can steadily creep into your marketing strategy if you don’t have a strict plan. Companies with multiple people working in the marketing department need to assign specific tasks and roles. They can’t risk someone inadvertently contaminating the testing conditions by adding extra variables.

You want accurate results, and the only way to get accurate results is to make sure you have a plan at all times.

The Duration Of The Test

There are no rules for how long you should make the test last. This varies from company to company and from variable to variable. What you must never do is have it last for a short period of time. There’s no real danger of making it last too long, other than inefficiency.

Shorter tests, though, will prejudice the results. If you only gain 10 visitors and they all respond to your new variable, it would be wrong of you to conclude this variable will drive your company to new heights. These 10 visitors might have given you an uncanny stream of luck.

The next 10 visitors might not respond to any of your changes. But you don’t know this because you haven’t given the test enough time to run.


Throughout a period of prolonged testing, you need to make it as efficient as possible. You don’t want to lose days and weeks dithering over little details. This goes back to the point about having an action plan in place.

Firstly, test the control. This is the site with no changes made. It’s the base value. Pinpoint your control. This could be a simple call to action, or a specific banner. Your control must remain in place for every test, regardless of the variable.

If you have an action plan from start to finish, you will finish you’re A/B examinations on-time without any delays.

Risking The Flow

The problem with a lot of companies is they won’t allow a test to run for long enough because it risks those precious, precious leads. They’ll give you perhaps 10 or 20 per cent of the traffic flow to conduct your tests, before referring back to how the site was before.

Bosses like to use what has always worked, so the experimental is something they’ll look to avoid. You have to break this way of thinking. You can’t get accurate tests by using such a small sample size.

Yes, major changes are a risk, but this is why it’s best to only change one minor variable at a time. This takes away the ‘shock factor’ and minimises the risk of turning away new and existing customers.

Get The Public Involved

Bickering with your boss about what you should do first is only going to lead to lost productivity. You shouldn’t waste your time with something like this. Instead, compromise by getting some opinions from the general public. Offer a discount on some of your products in exchange for them filling out a survey.

Ask generic questions you can glean an answer from. Ask them what they would prefer to see on your website, and what they’re already happy with. By allowing your audience to dictate what you do, you’re more likely to have success with your testing.

You should use analytics to work out what you should do first and the potential benefits of going in a certain direction.

Tie It To The Payoffs

Risk is a part of marketing. And it’s a part of marketing you won’t be able to remove. You can only minimise the risks. Unfortunately, companies can’t handle this. They want to see payoffs right now, and they won’t accept a huge risk tied to a payoff which isn’t worth it.

With your A/B testing, you have to provide a worthy payoff for the risk you want to take. Think of the potential benefits of changing the makeup of a website, such as more of an ability to engage online customers, more conversions, and increasing SEO rankings.

Overall, A/B testing can prove a worthy ally in your fight to get more conversions. It’s not as simple as it sounds, though. Making minor changes to the way you approach this sort of analysis can have a big impact on your results. Spend time planning your testing campaign and you’ll reap the rewards later on.

Richard Myers is an app developer, tech buff and an overall geek. Years of working with servers of all kinds has endowed him with the knowledge that he now likes sharing with anyone who needs his help. He often uses secure hosting solutions offered by Singlehop.