Table of Contents
Search Engine Optimization is a skill that is constantly changing and evolving. If your business depends on targeted traffic, then resting on your laurels and sticking with what you know simply isn’t an option anymore.
It was only a few years ago that we believed the best way to succeed at SEO was to stuff a page full of keywords and to pay for links from as many sites as possible. Of course, we know that this will get a site penalized and even de-indexed quickly these days!
Today, Google is an AI-driven company. Google wants to combine its search engine with its smart assistant, so that you can simply ask Google who was in X film, or how to cook X meal, and it will be able to answer in natural language by pulling from the web.
Future implications of this aside, what it means for webmasters right now, is that they need to start being smarter about their use of keywords and the type of content they produce. They need to think about using markup language, employing latent semantic indexing, and getting links from sites that Google trusts, with plenty of authority. They need to do all this while STILL producing content that is engaging and entertaining for real, human readers.
This is such a nuanced and complex topic, that you’ll find you can no longer just ‘guess’ what will work. And you can no longer make assumptions.
The ONLY way to succeed online, is to take a data-driven approach. That means watching your stats and metrics very closely, but it also means running split tests, which can help you to identify the precise techniques and strategies that are guaranteed to bring your website success.
In this article, you’ll learn why this is no longer optional for SEOs, and why it’s crucial for the entire industry that more agencies start using this topic. You’ll also see precisely what an AB test is, and how you can employ it to amazing effect.
So, just what is an AB Test?
AB Tests, also known as split tests, are tests used in marketing and business to ensure that a strategy works before going ‘all in’. They do this by testing that strategy in a real-world scenario, using controlled conditions.
Essentially, this works just like experimental design in research. That you will split your test subjects into two groups: an experimental group and a control group. You’ll then attempt to keep as many variables (factors) the same as possible in the two groups, change just one thing, and then compare the outcome to see which worked better.
So, an example of how this might work in business, would be to think about a fast food chain that currently uses a bold, sans-serif font for its name on the front of the buildings. The owner of the business conducts some market research, however, and learns that customers are not too keen on that logo, and would rather see a more ornate font to help welcome them into the building.
What does the owner do? One option would be to change the text across all stores and to hope that the market research proves to be accurate.
What Exactly Is An AB Split Test?
The business owner did something right here, in that they at least conducted market research first, which helped to motivate a smart change. The worst case scenario is that the business owner would simply decide that they like a more ornate font and change it across all their stores without thinking about the potential implications. They could end up losing a lot of money and never knowing where the ‘leak’ was.
But the problem is that even with market research, you still can’t be certain that you’ve made the right decision. Market research can be wrong, and it can miss other factors (the visitors may prefer the more ornate font, but they might not see it as easily as the clear, bold font!).
A split test then would involve changing the font on half the buildings. It would then compare the profits of those two groups, in order to see which one performed better and whether the change should be employed across the entire chain. If the stores with the new font earned more money, then the same technique could be used globally. If not, it could simply be forgotten. And another new change could be tested.
This way, split testing allows you to be certain that something is the right move before you do it. And the same thing works just like that in SEO. Here, you might test changing keyword densities, or you might test using more headers. Maybe you test using more images, or perhaps you try using an image at the top of the page.
All these things might end up impacting on the performance of your pages in the Search Engine Results Pages (SERPs), and thus help inform your practices in the future. You might also go back through all older pages on the website and make the changes across the board.
How do you run a split test for SEO? One option is to take lots of similar web pages that are all performing equally well, and then to make changes to half of them. Head into your Google Analytics account and make a projection for how you would normally expect them to perform, and then compare that projection to how the pages actually perform.
If all the pages with the changes start performing better than projected while the other pages remain the same, then you can take the changes and adopt them permanently on all versions of your site. If they start to fall in the rankings, then you can drop the change and move onto your next strategy.
Another option is to use a redirect – creating two versions of the same web page and sending half of your visitors to each. This will allow you to conduct a more scientific study, but it will also mean that you can’t measure your position in the SERPs (seeing as it’s just one page and Google will only rank it once). This can nevertheless be used to see how changes affect things like bounce rates, CTR, conversions etc., which all are likely to have a knock-on effect on SEO AND on your profits.
But Do I Really Have To Split Test?
I know what you might be thinking at this point: do you really need to go to all the trouble of creating two different versions of the same pages? Or split your pages into groups that way?
Couldn’t you just introduce the changes to your site and then see how they perform?
Or perhaps just test the changes on one page?
Maybe just read a book??
Well, the problem with the first option – introducing the changes to the site – are several. Firstly, this would take a long time. If you go to all the trouble of introducing changes to all of your pages, only for this to have no beneficial effect, then you’re going to have wasted a huge amount of time that could have been spent much better.
Not only that, but if you make all these changes to every page on your site and the changes turn out to be a mistake, then this can seriously hurt your business. Not only can this hurt your short-term profits as every page on your site begins to underperform, but it can also seriously hurt your reputation if those changes end up driving away your visitors.
Okay, so making changes across the board isn’t an option. But what about introducing changes to just one page and seeing how that works?
Well, the problem in that case is that you’ll be introducing what are known as ‘confounding variables. A confounding variable is anything that could skew your results but that you haven’t accounted for. So, in this case, if you include your new SEO strategy in just the very next blog post, you might see it thrive and assume that means it was a good move. You’d then make the changes to every post on your page.
But what if the only reason that page did so well is that you happened to write a good post that your audience loved? Or what if you accidentally stumbled upon some other SEO-secret sauce. Maybe your headers were all the perfect length, and it had nothing to do with the resource box at the bottom of the page?
In this case, you could once again go to the effort of changing every single post on your site, only to find that the strategy doesn’t work, and you just seriously damaged your website ranking and your reputation. Bad news!
Finally, why conduct these experiments at all? Why not just use the knowledge provided by the SEO community – or even other people who conducted those experiments?
What Split Testing Has Already Taught Us About SEO
A great article on Medium, shared by the creators of Optimization Delivery Network, explains this very clearly.
ODN provides SEO services for clients, and the company recently decided to try using split testing to make better choices (it developed an Optimization Delivery Network for this very purpose).
The company then began to conduct hundreds of experiments on a huge number of blogs, to collect as much data as possible. So, what did they find, and what can we learn?
The first thing they found, was that using a more popular search term doesn’t always yield better performance in the SERPs.
Let’s say that you have a web page that ranks well for a term like ‘how to earn a living online’. You then use your keyword research tool, and you see that ‘make money online’ is a more popular search term that has a higher volume of searches.
Your obvious next step, therefore, might be to start trying to rank for that term instead – and so to swap the search term.
As it turns out though, according to ODN’s research, this doesn’t always work. In many cases, choosing the more popular search term hurts rankings and results in fewer views. There are many potential reasons for this, perhaps it results in poorer natural language. Or perhaps the use of more ostentatious keywords hurts user engagement.
Whatever the case, though, the reality here is very much subverting the expectation. So, it’s a good thing they checked!
Another thing the company learned, was that 30-40% of ALL the advice typically suggested in an SEO audit has zero effect whatsoever. So, if you were to conduct an SEO audit for a client, make a note of all the things they aren’t doing quite right, and then advise them to make a bunch of changes… what might happen?
Well, you’ll meet resistance. The professionals that work for that organization will carry out all the changes as they are instructed, but they’ll likely find that it’s an arduous task that seems pointless. They’ll grumble, but they’ll do it. And the company managers might also groan that their staff are spending all their time on these fiddly tasks.
So now imagine that after all this: after all this seemingly pointless work… it turns out that it was pointless! There is no benefit in the SERPs, and the site stays put exactly where it previously was. This could seriously undermine your authority, hurt your reputation in the eyes of the client, and possibly result in the company no longer using your services.
In fact, this doesn’t only hurt your business – it hurts the entire SEO industry. As it is, there is already a fair amount of bad press surrounding SEO.
So, when you carry out that SEO audit, you should also carry out a split test or several. This way, you can then demonstrate to the client precisely why they should be following your advice, and what will happen if they do. At the same time, you’ll be able to ensure that they see the results they want to see.
Do You have to Keep Doing It?
One last question you might have: do you need to keep doing it?
Once you’ve found out what works and what doesn’t, can’t you just leave it at that?
The problem with this idea, is that you’ll find the best advice changes from website to website and even over time.
For starters, Google is constantly changing its algorithms and updating the way it trawls the web and ranks its content. What works right now will not necessarily work in a month’s time.
At the same time, though, what works for one website won’t necessarily work for another. In one study, it was found that the same strategy would improve rankings by 6% on one website, but would lower them by 3% on another.
This might be due to differences in the niche, the way that those fans reacted, or how that change interacted with countless other changes on the site.
So yes, you really should keep split testing. In fact, any time that you decide to make a big change across multiple pages, you should perform a split test.
The good news is that this doesn’t have to be a lengthy process, and with the right tools and techniques, you can run tests quickly and efficiently.
Okay, so split testing isn’t exactly a way of life. But it certainly is much more than just a single technique. Split testing is a certain way to go about things that can have a hugely beneficial impact on nearly any aspect of business.
The simple notion, at the core of split testing, is that you should test everything and act in a data driven manner.
As mentioned, the worst thing that a business could do when it comes to the font on its stores, is simply to change the font on a whim based on the owner’s gut feelings. That might not seem damaging, but ultimately the store isn’t designed for the owner… it’s designed for the customer!
Why go to all that effort on a hunch?
One phrase that is often used in business that has a lot in common with the concept of the split test, is ‘fail fast’.
Split Testing Is Actually A Way Of Life
Fail fast doesn’t mean setting out to fail quickly – rather, it means setting out to fail OR succeed quickly. If you are going to fail, then it is better to do so fast.
One of the biggest mistakes I often see people making, is coming up with a new idea for a website business, and then investing months and months into its creation. So many people tell me they’re working on a business idea, but they can’t tell me what it’s about (obviously, for I am constantly poised to steal their ideas since I have nothing of my own going on!). Often these are people with no jobs, still living at home with their Moms. Maybe they’re building the site with a friend.
When the business launches, the creator has no choice but to succeed quickly. After all, they haven’t been earning anything for years, their parents are extremely close to kicking them out of their home, and they probably took out a business loan that needs repaying.
But very often the company doesn’t take off in a big way, and they end up quitting and going back to the drawing board. No one clicked on their site, it didn’t rank, there was too much competition, and there wasn’t a demand for it.
They just wasted months or even years of their life. Ouch.
So, what should they have done instead? Designed a simple and easy business model that they could put together in a few weeks. Then release it. If there is an audience for it, they’ll see it start to succeed right away. If that happens, they invest more time and effort into making it perfect.
If not? They ditch it and move on to the next idea – no harm done.
Better yet is to use verification – to ensure that the business idea works. You could do that by trying to collect pre-orders, by testing the keyword you’re hoping to rank works by using Google Ads to get to the top of the SERPs for a few days, and by looking at how other sites perform.
All this means acting in a data driven manner and looking for evidence that your idea has legs before you invest years of your life into it. That is all split testing really is, but on a much smaller scale.
Split testing means taking every big decision and change you make on your website and testing that it will help your success rather than hurt it. That means never wasting time on fiddly SEO that does nothing, and it means gradually improving your site bit-by-bit every day.
Split testing is a fool-proof business tool in that way because it is determined by fact. Don’t guess what might work, prove it. In doing so, you’ll boost your own success and even help the SEO industry.
Of course, there are a few downsides to split testing, a few pitfalls, and a few ways it can often go wrong. Before you dive in then, you should familiarize yourself with all of these and spend some time better understanding the concept.
In the video course, you’ll learn to become a master of split testing. You’ll discover powerful tools, lesser-known advanced techniques, and best practices. All of this can help you to make much smarter, informed decisions for your own business and for your clients.
