What is A/B testing?
A/B testing, also referred to as split testing, is an experiment where two or more versions of the same webpage are shown to visitors. A statistical hypothesis is then used to determine which one performs or converts better. Both variations A and B are shown randomly to users, with each version’s performance analyzed based on clicks, purchases, and subscriptions it gets.
Why You Should Conduct A/B Tests
A/B testing allows you to get a complete handle on what parts of your website are drawing in the best custom. By testing out each aspect of your site, you can see where the majority of your traffic is coming from. It can be expensive to pay for an advertising campaign that covers every aspect of your online store. With A/B testing, you can see exactly where to focus your spending on maximizing your profits.
A/B testing also allows you to measure your customers’ experience with your store. By learning how good or bad your customers find your store, you can alter how it works to suit them better. This can be through simplifying the buying process or just limiting the amount of unnecessary exposure they experience. In short, through A/B testing, you can revolutionize your store to make it easier, faster, and more enjoyable for your customers. If your customers enjoy their time with you, they’re much more likely to come back again and again.
Types of A/B testing
Each aspect of your store has its type of testing. As you go through your testing phase, you will need to consider each aspect, such as:
1. Element-Level Testing
This is a very simple set of tests designed to explore how effective each element of your landing page is. It will allow you to test:
- Calls to Action – test variations on font, color, location, and wording to see what most effectively prompts customers.
- Headline Testing – as the first thing your customers see when they reach your landing page, it is important to find the most attention-grabbing headline you can. The more effective it is, the more likely people will follow through from Google Ads, email-open rates, impressions, and so on.
- Visual Content – Everyone knows pictures are integral to the online selling world, but at what point do you have too many? Or too few? Can you replace text with videos, audio recordings, or infographics? Or perhaps your content needs more words and less imagery? These things all need to be tested.
- Product Descriptions – knowing just how much information your customers need before they buy an item is vital. Too much can lead to them abandoning their cart, and too little will prevent them from buying it at all. Test the length and depth to which your descriptions explain your products to find the perfect middle ground.
It can be a slow process to go through each area individually, so it will pay to be thorough. You don’t have to test each one on its own, so keep a running strategy in mind and test multiple elements together to find your winning combination. Don’t be discouraged if every tiny change shows little difference.
2. Page-level testing
Otherwise known as multivariate testing, page-level testing allows you to test two different layouts at a time. By setting up two different page layout styles, you can determine which elements were most effective at converting sales.
The overall design of your landing pages is vital to conversion. It is speculated that a great landing page can improve sales by between 3-5x. It also gives visitors the chance to show which type of page they prefer. After all, some customers prefer to read as much about a product as possible, and others just want the basic information. You are determining which type of person your target audience will help you plan out the rest of your site.
It should be noted that this type of testing is only effective for pages with a high level of traffic, and it is difficult to derive meaningful data from smaller test pools in this way. Therefore, if you are just building your audience, it will be better to save these broader tests until later.
3. Visitor flow testing
This is a much more complex type of testing involving modifying the routes your customers will take on your site. It can not be very easy to rearrange the layout of your website, but it is the most rewarding if done correctly.
Through visitor flow testing, you can see whether people are more likely to convert when taken directly to a sales page or if they prefer to see other parts of your store first. This can be done by including pages for alternative offers, similar products, or bundle offers as a step between the landing page and the checkout. You can also include pages with further information regarding the products customers are looking at.
Compare this data with a more simplistic model that sees your customer going from the landing page straight to checkout to determine best what suits your audience.
4. Overall strategy testing
While you are testing the various aspects of what your customers like on your site, it is also important to test how effective getting them there is. Overall strategy testing involves you studying what parts of your advertising strategy are most effective. This includes:
- SEO A/B Testing – looking in-depth into the various SEO strategies you are running will help determine the most effective. If certain keywords are getting you nowhere, try changing them or swapping them for others.
- Social Proof Testing – there’s no better draw for customers than the testimony of other customers. But which ones are the most successful for your company? Do trust pilot reviews draw more sales or Facebook comments? Are you picking up more traffic from social media mentions or influencer plugs? Test all these things and find which one most helps you.
Where you should do A/B testing
As every online business is different, you cannot expect the test or results of one company to reflect the needs of another. A store selling sports shoes will not have the same audience as a bookstore, so the methods used to draw those audiences in and what is most effective for them will differ vastly. This does not mean that there aren’t some areas that you can test which are universal.
Aside from testing your data, it would help if you also tried some of the following tests:
Types of A/B testing
It is not a good idea to jump straight in with a free shipping policy, and it can be used to test conversion rates. By adding an A/B test for free shipping alongside paid shipping, you can see how effective your standard selling procedures are.
Shipping costs are a powerful factor in cart abandonment. If you find that your customers are never quite hitting the complete transaction button, it may be time to try out a few free shipping cases. If this doesn’t have any effect, then you know the issue lies elsewhere in your store.
It is best utilized by creating a threshold value on your items – a cost at which your potential customers can earn free shipping if they spend enough. Sometimes, an option for “Free shipping on purchases over $40” can be the difference between decline and profits.
Calls to Action
There are so many different ways to display your CTA that testing can take months to get the perfect one. But this is time well spent. Try experimenting with different wordings, fonts, sizes, and colors. Try testing whether having a CTA at the top of the page works better than at the bottom. Perhaps customers are already interested in your products but get put off wading through a whole page of text.
The way you word your CTAs is also important to consider. Some people are put off at the prospect of “sign up” but will happily “join now.” Others might be more inclined to click on a “free trial” over a “get started.” “Buy Now” has a greater sense of urgency than a plain “order here” will.
Test multiple buttons on your site to see which draws the most attention. Two buttons of different sizes, colors, font, and wording are ideal for determining which kinds your customers prefer and can be easily measured by click-through metrics.
Headline wording
It’s often not what you say, but how you say it. The wording of your headlines matters far more than you’d realize if you weren’t testing for it. Simple, honest statements can easily get noticed and shared on the internet, while more detailed ones might convert quicker.
Using as much variation as possible while A/B testing your headline wording will help you determine what spreads your company best and inspires sales. You may discover that a handful of words work best in all situations, or perhaps a specific brand phrase calls fans running.
Studying the click-through rates, bounce rates, reading times, and shares will help you determine what is working and what is not. You can also get ideas for what works best by performing a few Google searches. Whatever words and phrases that pertain to your business show up most often reveal what is most effective. Never forget to utilize this tool.
Typography
Not just choosing which fonts, colors, or word sizes you use, the typography of your pages also includes aspects such as word positioning and blank spaces. Remember, the overall layout of the words you choose also leads the customer through your site. Some companies prefer a sleek and minimalistic approach, choosing large empty spaces with small fonts to give their audience a clear view of the majesty of their products. Others like to provide as much information as possible, providing a sea of text to buoy up a complicated product.
Look around at your competitors to see what best suits your business, and then test variations of this style on your audience. Finding that perfect balance that provides your customer what they need to know alongside what they want, and giving them a comfortable buying journey along the way, will boost your conversions and your profits.
A/B testing mistakes to avoid
A/B testing can be a waste of time and money when not done correctly. Here are a few common mistakes to avoid with your split tests:
- Testing pages that don’t affect conversions: Split testing web pages that don’t have a direct impact on conversions isn’t going to do you any favors. With limited resources and time, it is essential to choose the best candidates for testing.
- Testing multiple elements in one go: If you run tests with multiple elements simultaneously, you will have no way of knowing what is bringing results. This will negatively affect your chances of forming solid hypotheses. You will also land with less-than-optimal results for the pages on which you ran tests.
- Using small sample sizes: If you do not consider the rules of good data science – calculating a sample size of 80% and 95% statistical significance – your results will not be satisfactory. It will not change your goals in the long term.
- Borrowing ideas: Doing competitor research and using case studies to create an informed hypothesis is a good idea. It becomes a mistake when you are only getting ideas from these sources. Many of your best conclusions will come from tests that even your competitor hasn’t performed.
- Testing seldomly: Split-testing is something you do for as long as you have a website, and it is not a one-time effort. A strategy that accommodates changing customer behaviors involves sustainable testing over the long term.
- Creating hypotheses based on assumptions: Every A/B testing team will have one idea based on assumptions. An assumption will not qualify as a solid hypothesis because there is a lot of work to do. The point of A/B testing is to identify positive changes.
- Using analytics: If CTA’ B’ is converting at 10% and CTA ‘A’ only at 5%, it is the end of the story, right? No, test data includes a lot of information about your users, like high-converting segments, on-page hurdles, etc. Use an analytics platform like Google Analytics to get more out of your test results.
- Overlooking small wins: Retailers usually expect that there will be a massive result and do not pay attention to 2% or 3% changes. It is understandable in a way. But this is a mistake. Small wins are just as valid as large wins because tests with high statistical power can detect a small win and apply it to your strategy for great changes.