SEO is blessed with more tools and data than ever before. We have the ability to measure, track and report much of what we do, but are you making the most of this? I argue no, but more than that, I recommend that the industry doesn’t adopt the SEO testing mindset that helps to better understand the impact of change and truly improve and iterate. ..
The idea of testing is not to assume that change will be positive or negative. Rather, look at the data to explain the results. SEO with a test mindset doesn’t test once and considers it “completed”. Those who have the idea of testing expect to prove wrong, be prepared for the contingency, and react appropriately when it happens.
But what does this really mean? read!
First, there are some basic SEO truths you need to know. This isn’t just for testing, but for making (and monitoring) all sorts of changes to SEO.
- Any changes you make can impact performance
- Not all changes have a meaningful or significant impact on performance
- Other factors, such as algorithm updates, competitor activity, and even crawling and indexing your website, are constantly impacting performance.
- We do not fully understand the timing and magnitude of the changes behind point 3.
The last two points significantly reduce the reliability of many predictions. However, that does not mean that certain changes should not be investigated and tested. In fact, that means you should do it more often and more diligently.
You can make a strong hypothesis (more on that later), but the chances of unrelated factors being equal or better (and vice versa) at the same time are not zero. Therefore, even if your idea is good, a single test may not show it in the results.
How likely is it to happen? Well, it’s certainly relatively low, but if you’re serious about mastering SEO, it’s not something you can ignore.
Clarify what you are testing
In testing, size is important. The size / scale of the change has a significant impact on your ability to detect potential benefits. This doesn’t mean that small changes won’t give you big results, but testing small elements is unlikely to give you enough valuable data.
Are you deploying a few new title tags or migrating your domain? You can test either, but monitoring these two tests is very different.
When testing risk mitigation, you need to gather as much knowledge as possible. Therefore, it is advisable to test these larger changes for two reasons.
- Easy to measure effect
- If you really need to make these changes, they are much more risky
please think about it. Blindly migrating a domain without test data can have a serious negative impact on your website’s performance.
In addition, small, highly isolated changes can be difficult to control. If you change the title tag and notice that the ranking you’re tracking is increasing, is that decisive? It’s unlikely.
But does that mean you shouldn’t test them? Not at all. You need to duplicate the result multiple times to get accurate data.
A / B testing (detailed below) is the best way to control some external factors that make it difficult to interpret test data if planned correctly. However, if in doubt, try making as few changes as possible and then duplicating the test results over and over again.
SEO test method
The barriers to adopting the idea of testing are low and very low. Even performing sophisticated tests has never been easier.
Below are some of the main ways to focus on this.
- Tracking / Annotating Deployments for Key Indicators — The lowest standard of testing, yet not done by many. At a minimum, you need to annotate visibility data, Google Analytics, or another workflow spreadsheet to highlight changes as they occur. The day / week / month results provide a benchmark and can be reverted to this in the future.
- Time-based testing-If you have already set up the Google Search Console or web analytics package, time-based testing does not require any more important tools or overhead. Effectively compare (or run year-on-year) the pre-test and post-test periods. This doesn’t explain the externalities well, but if the post-tests differ from similar periods, this can tell you something.
- A / B (split test)-A / B testing is the gold standard for testing, but may require additional overhead or planning. In this scenario, we present Google with two different versions of page templates (entire test / control buckets) to index and rank the differences. Learn more about the process here.
Report tests and track metrics
Obviously, I want to track the increase in revenue as a direct result of the split test, but just setting these as test metrics causes a lot of problems.
You need a set of metrics, and you need to understand how they affect each other and how your tests affect them.
- impression-Although it’s a partial vanity metric, impressions are a key indicator of clicks. Even if your tests didn’t increase your clicks / traffic significantly, they may still indicate an increase in your ranking.
- Clicks-Increased clicks are clearly an important indicator and it’s most important to measure this
- CTR (click rate)-CTR can be useful, but you need to be very careful. If you’re tracking SERP features, it can be important to understand CTR changes.
- Ranking-Ranking is another key indicator of performance. This is a plus if you see an increase in the ranking (or visibility score from the ranking). A common mistake people make is to assume that rankings are directly related to traffic. If it’s not on the first page, it won’t work, but that doesn’t mean the test didn’t work.
- (Organic) traffic-It’s similar to clicks (not the same!), But it could be one of your business’s most important metrics. If you can prove the increase here, it will be the strongest support for the test.
- (Organic) Profit— This is a metric set award, but the relationship between ranking and revenue is not as interrelated as people think. If you’re driving clicks / traffic, but you’re not converting due to low inventory (for example), or your page speed is slow, you won’t see any significant revenue growth.
We need to give these metrics a sense of proportionality / importance. Some are obviously more useful for business than others. However, these goals depend on what you need, so we cannot explain how to evaluate individual indicators.
If your tests show only an increase in impressions, the benefits are relatively low. But if you can track sales growth, there’s something serious to celebrate.
Creating a test hypothesis
There are many ways to hypothesize a test, but changing the blank input in the next statement is a good starting point.
If i [make this change] Will happen [positively/negatively] Impact [this/these metrics] To [x%] upon [y pages] because _________
This may seem daunting to export with each change, but doing so forces you to:
- Know exactly what your changes are and what pages are affected
- You can do this for changes that you think will be negative, whether you find them useful or not (!)
- The main measurement method needs to be planned
- The impact needs to be demonstrated. Additional investigation may be required to do so in a reliable manner.
- Finally, we need to show why we think this is the case.
Data must be collected and analyzed to fully draw meaningful conclusions about this, but the process itself makes it easier to consider actions and report / reflect results.
SEO test lessons
Even with the idea of testing, SEO is complicated. At first glance, the test provides some answers to some “big” questions, but this does not mean that you can be satisfied / lazy with what the test offers.
- Running tests can be time consuming and the data can be inconclusive. Therefore, plan your test properly and get approval first.
- Testing to “solve a dispute” does not make good use of someone’s time, and a positive result does not always attract the attention you believe it deserves.
- Negative tests can provide a lot of learning
- The less traffic, the longer it will take to run. In some cases, the usefulness of the results is ruled out.
- If you don’t receive a lot of non-branded traffic, the impact of your “brand” activity can easily invalidate the profits / losses made in your tests.
- If you’re not running split tests, or time-based tests, you need to consider seasonality.
- Tests show the results of making certain changes in a particular environment at a particular time. Do not assume that these findings are portable to all other scenarios and expect the same results.
What are you looking for? Make a hypothesis and start writing a framework to test your last SEO change. If you’re already used to this, it may be time to consider more advanced testing methods to prepare your business case for larger and more meaningful testing in your business.