How To Run An SEO Experiment

Blog in my soup July 12, 2013 0

In it’s own way, SEO is kind of like science. SEO is a constantly changing landscape that we learn new things about each day. It’s all about theories about what works and doesn’t work, and how that changes on a day-to-day basis. Because of this, one of the most valuable tools at the disposal of SEO specialists (as well as all scientists) is to run an experiment. Now for those of you who haven’t taken a chemistry course in a long time, we’ll give you a refresher on the steps of running an experiment.


1. Identify a Problem/Something You Want to Test

If you’re going to run an experiment, you have to have something you plan on testing out or “experimenting” with in the first place. Maybe the conversion rate on one of your web pages isn’t that great. Or you’re curious whether someone giving a +1 on Google has an impact on your ranking on the SERP. Whatever it is, you need to identify a potential problem or issue that you want to address or find out more about prior to beginning an experiment.

2. Formulate a Hypothesis

Consider your problem. What do you think might solve that problem? I.e. Begin to come up with a testable theory that might explain or help you better understand your problem. These usually take the form of “If, Then” statements. For example, “If a site receives more inbound links from an external domain, then it will rise in PageRank.” This is a (very basic, and proven to be true) theory that can be tested by an SEO specialist in order to have a better grasp of how linking functions in SEO and search engine algorithms.

3. Test Your Hypothesis

This is where the experimentation part of the experiment enters. Now you will need to test your hypothesis by conducting an experiment. Usually this is done by creating a variable domain or web page. As with most experiments, you’ll likely want a “control group” which will probably be your original web page, to make sure a dramatic change in your ranking comes from variation within the experiment and not some outside cause. (For example, if it’s discovered that toasters cure cancer, and your site sells toasters, that might cause an influx in traffic to your site or a higher conversion rate that was not the result of your experimented variation.) Some of your results can be trackable by a simple search on Google (if you’re checking for rankings on the SERP) while others might require you to use more specialized tools such as Moz’s OpenSiteExplorer or GoogleAnalytics. Google also offers a Google Experiments API which allows you set goals and test metrics on several variations of a page or domain. You may want to decide upon a set amount of time for your experiment; longer experiments can be more reliable due to an averaging of traffic and seasonal purchasing patterns, (e.g. swimming pools are a more popular purchase in the summers, searches for bars and clubs are more popular on weekends), but also obviously take longer to track.

4. Collect and Analyze Data

If you’re using a service like GoogleAnalytics or the aforementioned Google Experiments API, then your data will be collected automatically. If not, you may need to obtain the data yourself. As mentioned previously, you can do certain data collection such as checking search engine rankings manually by doing a regular check on how the domains are performing on Google or another search engine. (Remember to be consistent in keywords though!) Once you have the data collected, you’ll need to analyze it. Usually this amounts to a simple comparison between your variations and the original page to see which one performed better for the metric you were testing, but can be more complex if you test multiple variables or outputs.

5. Reach a Conclusion

The final step is to conclude something about your SEO experiment based upon the data you have collected. In a traditional sense, this is making a statement about your hypothesis—specifically, affirming whether it was true or false. However, you may be able to conclude other things about the experiment that you weren’t testing. Maybe the variable you altered caused an unexpected change. Use your results to formulate a conclusion about the experiment and decide whether to alter your SEO practices accordingly. However, no experiment is 100% accurate, and experiments frequently can and do suffer from human error as well as other unforeseen factors. The best way to ensure a conclusion is absolutely correct is to perform multiple experiments (though time and budget constraints don’t always allow this). All in all, be vigilant and if you implement a new strategy that was successful in an experiment, be aware that your strategy may not replicate itself exactly. Algorithms change, mistakes are made, and sometimes conclusions are flat out wrong. SEO is constantly changing and the best defense against that fact is to always keep an eye out and quickly adapt to those changes.

Ricardo Casas is the CEO and founder of Fahrenheit Marketing LLC, an Austin, Texas based internet marketing company. With over ten years of experience in internet marketing, Casas knows the importance of understanding SEO and the knowledge that can be gained by SEO experiments.
Is your Domain Still Available? Use the TransIP Domainchecker!

Enhanced by Zemanta

Leave A Response »