There’s a massive amount of things you’re “supposed” to do getting your affiliate site’s visitors to click and spend more money. It’s pretty much impossible to know which conventional advice really will work and what is complete hogwash.
To make matters worse, what you might not know is that some conventional advice can actually cause you to lose earnings! Sounds terrible, doesn’t it? How can you be sure you’re on the right track when making changes to your posts? What can you do?
That’s where the principles of conversion rate optimization (CRO) and A/B testing can help. Not only that, it’s a great way to get more earnings from the site traffic you’re already getting.
The great news is that you don’t have to have a huge site to take advantage of it. Even many small sites just like mine can put these simple ideas to work right away!
What is conversion rate optimization?
Conversion rate optimization (CRO) may sound super-complicated or like something you shouldn’t worry about. However, that couldn;t be further from the truth!
Let’s start with the basics:
CRO is a method of finding ways to test changes to your site & posts and discover ways to increase (optimize) earnings. That’s it in a nutshell.
We’re mainly worried about actual conversions, not just clicks. This is a very important distinction I’ll explain later.
The scientific method approach
It’s the way that we go about doing so that’s so critical. Unlike guesswork, the principles of CRO require that we do the following, much like the “scientific method” you may remember from school or college:
- Create an idea (hypothesis) for a change we can test: This change (test variant) might increase affiliate clicks & earnings or it could perform worse
- Test the hypothesis via A/B testing
- Collect data
- Use the data to reach a statistically significant conclusion and find out if our hypothesis was good or bad: Did we improve clicks and earnings?
As Kurt Philip from the professional CRO agency Convertica taught me in his CRO Academy Pro course, what works on one site might not work on another: We have to test everything!
Data is EVERYTHING!
Anything that someone claims is “better” or “increases conversions” means almost nothing without cold, hard data. Without data-backed proof, you will never know if you’re missing out on clicks & earning improvements that could be easily gained or if you’re being misled.
A/B testing is where we find out what works and what doesn’t. In other words, just like your school science lab, it’s where we carry out testing our “science” (CRO) experiment.
How does A/B testing work? The basics explained
A diagram showing the basics of A/B testing. A testing application (one of several types: plugins, added code used with a service, or other) controls which version of a post under test is seen. Their click-through rate (CTR) or other test actions are recorded and a statistical analysis is performed. For accurate CRO testing, the conversion rates are also tracked using unique test IDs.
A/B testing is actually fairly simple when it comes down to it. That’s because the real work is handled by a test application which can be one of several types:
- A simple plugin
- Service-provided plugin with a service provider’s data services (subscription, paid service)
- Code added manually to posts then tracked using an external service provider.
In all cases, the principles are the same. For A/B testing we want to compare an original post (“A”) vs our test version (“B”) which has an item we want to track behavior for.
Therefore we only use one change we’d like to test at a time.
The application controls which visitors arrive at which post and also provides a way to create the test post. Visitors arriving at post version “A” aren’t aware of post version “B” and vice versa.
The “conversion” or “action” tracked by the user is chosen during test creation. Typically this is an outbound affiliate link leading to a sales page.
Data tracking & conclusions
Data from visitors (most often the click-through rate [CTR] ) is recorded and evaluated by the application’s algorithm. Periodically the results are updated and the site owner can see what’s happening up to that point.
After enough visitors have been tracked (usually around at least 1,000 in most cases) the algorithm will reach a fairly good confidence level (statistical confidence in the results) for the test.
Testing takes time & sufficient traffic
Testing does have a drawback: It’s definitely not a fast process.
Another thing I need to share is that it’s critical to have a sufficient amount of visitors to your test posts. More often than not you’ll need at least 1,000+ visitors before a reasonably good conclusion can be drawn. The more the better, too.
As Kurt Philip would advise, a statistical confidence level in your results is about 90%-95%, with 95% being ideal.
For a rough time estimate, divide 1,000 by the number of unique monthly visitors you have for a post you’d like to test to know how many months it will take to test that post. If a post has less than 500 per month it might be better to wait until your site has grown and has reached 10,000+ visitors per month total.
In my test experience (with about 15K-20K visitors per month to each of my sites) it can easily take 1-2 months or more to get confident results for an A/B test. Anyone who starts a test then ends it too soon because they think their test change is getting great results is making a huge mistake!
That’s because as the data sample size increases (more visitors) the algorithm’s computations will “average out.” You’ll often find that it’s much more important to look at the big picture: The overall long term results are what matter and not just a few dips or peaks in CTR.
However, CTR tells only part of the picture.
Clicks do not always equal conversions!
As I mentioned earlier the click-through rate tells only part of the story. In fact that’s one of the most common errors people make: They often assume that a higher CTR means more earnings.
As counter-intuitive as it may be, increased clicks do not always results in better earnings. The only way to correctly gauge a change’s impact is to tracking sales conversions.
Conversions can be tracked in many cases by creating additional “sub IDs” (test tracking IDs) within your affiliate account. For Amazon affiliates this is easy as you can create quite a large number as needed.
I use the following method: For each site’s account I create a series of tracking IDs just for this purpose with sequential numbers added to the end.
For example, if my example site’s main ID is “cathmk-02” I do create the following: cathmk01-20, cathmk02-20, cathmk03-20, cathmk04-20, and so on.
To be consistent during testing I use the odd-numbered ID for the original post under test (A) and the corresponding even-numbered ID for test post B.
During test setup I replace the original post’s tracking IDs with the test ID and likewise for the test variant B as well.
Real world examples from my own sites
Now that you’ve got a better understanding of the what, whys, and hows, how about some real examples?
Here are 2 of my actual CRO tests and their results along with the actual conversion data also.
Test hypothesis: For a budget keyword buyer post, product tables with prices shown should convert better than the original without pricing.
Image #1: Test results based only on CTR rate. As you can see, the original (no pricing) actually, on the whole, resulted in more clicks to Amazon than the alternative (with pricing).
Image #2: Sales conversion data shows very little difference between the two test versions. We can conclude that the hypothesis is proven wrong.
From the images you can see here, I based test results not just on the CTR (shown in the Nelio A/B test dashboard) but actual Amazon conversion data & earnings. From the data we can see that on the whole, the test results are that the original product tables performed marginally better than tables with pricing included.
Result: The test hypothesis was not a winner. This is not a CRO optimization change I should use.
Test hypothesis: Using better optimized product tables in buyer keyword posts (test post B) will result in higher CTR and better conversions (earnings) vs the original tables (A).
Image #1: Implementing improved product tables resulted in a significant improvement in CTR. But what will the CRO results be? Will my test hypothesis be proven right?
Image #2: The numbers don’t lie! Great news – my hypothesis was true. The optimized tables (test variant B) proved to generate not just a higher CTR but earnings as well vs the original tables.
As you can see above, this was a great example of what’s possible with CRO methods. In this case I used custom (optimized) product tables and wanted to prove for myself that they would help my sites generate increased clicks and most importantly, earnings.
After about 1,000 views during testing there was a statistically significant conclusion, backed by my Amazon test tracking IDs: The optimized tables performed much better.
I got a great boost to both sites I’m still using to this day.
While an improvement like this doesn’t happen every day (often times improvements of 5-15% might be more common) it’s a great real-world example of what’s possible.
To my surprise many people completely ignore CRO and testing. If you do so, you’re leaving free money on the table and costing yourself a lot of great experience, too!
CRO isn’t hard and it’s fantastic way to utilize the existing traffic you have to earn more money by trying new things backed by real data rather than guessing and actually losing earnings instead.
Guess work doesn’t work. You have to test and get real data. Testing properly means following the right process and a few basic rules but it’s fairly straightforward. Its an excellent opportunity for anyone with a relatively good amount of site traffic – even small site owners like myself.