Testing Google Algorithms with SEO

Google is constantly changing its algorithm to avoid people gaming the system, and this has been an effective strategy to keep website owners on their toes. Google is obviously reluctant to show people its algorithm, but there are many SEO professionals that are able to crack its code. How are they able to test the algorithm with accurate results? There is a science behind testing the algorithm, but it requires a lot of resources. You will need several servers, many domains, a lot of test data and you will need to make various changes to your websites to test the algorithm. If you want to run your own test, then follow the steps listed in this article.

Domains

You will need a lot of domains to properly test the algorithm. Google changes many things when an update is released, and you will need enough domains to test all of the differences. You will also need different types of domains to see if the changes affect domains of different sizes and topics.

If you want to run a small test, then you will need at least 50 domains, but most major tests are done with 150 or more. You should also have the domains in pairs for proper split testing. This means that you have two websites that are exactly the same with very similar domain names, or you test two very similar websites that are basically the same. The former is better, but the latter is adequate.

The domains should also be categorized by age and size. For example, Google might penalize a younger, smaller website more than an older, larger one. You can only test this by having various domains.

You should also put the domains on different servers. Having all of your domains on one server will dilute the results because Google usually penalizes people who have a large number of domains on a single server.

Servers

There aren’t quite as many requirements for servers, but there are a few you need to take into account before running a test. As stated above, you need to have your domains listed on different servers. There isn’t an exact science to this aspect yet, but you should only have about 10 to 15 different domains on each server.

If you want to test how Google has changed itself internationally, then you will need to have servers that are stationed in different countries. For example, the German version of Google will have different results than the American version of Google for the same search term. This is usually because Google favors local results, and it also uses language as a ranking mechanism. You rarely see German websites when using Google in America.

You don’t need to test the results from different countries if you don’t want to, but it will give you more comprehensive data. If you are only interested in domestic results, then just purchase server space in your country.

You should also ensure that the servers are similar. Google uses loading speed as a ranking mechanism, and a faster server might give some domains an unfair advantage. Unless you are testing this to see how much of a difference it makes, then you should try your best to use similar servers.

List Testing Attributes

What do you want to test? This is the most subjective part of testing the Google algorithm, but it is also the most important. You need to list all of the different attributes that you want to test. You also need solid, definable attributes so that you know what to look for when the test is over.

For example, you can test for the effects of thin content. Thin content is a major issue, and Google has been doing its best to drop websites that have poor articles. You can test the effects of thin content on both large and small websites, and you can also see how many bad articles it takes to destroy your ranking.

There are over 200 ranking elements that Google takes into account. You should try testing one or two elements on every pair of domains. This section is largely up to you. Just make sure that you have very definite test attributes in mind before proceeding forward. If you don’t, then your test may not give you any significant data.

Testing the Websites

Now it’s time to test the websites. You need to test the domains in pairs to see what happens when you make changes. The reason for this is because one domain will function as the control, and the other will function as the experiment. If you don’t have a control, then you won’t get proper results.

Make the changes that you want to test. For example, start posting thin content on the experimental domain. At the same time, post good articles on your control domain. You will need software to check your domain’s change in relevancy to gauge how important this ranking mechanism is.

Testing a domain is fairly easy, but you need to consistently work with the domain to get the best results.

Length of Testing

This is another subjective section because you can test the domains for as long as you want. At the same time, you should test the domains for at least six weeks. This is about how long it takes to get consistent results. You can test for longer, but that may not give you any significant data unless you are running a long-term test on your domains.

Conclusion

Running a test on Google’s algorithm isn’t difficult, but it does require a lot of resources. You should have many domains, and you need a good number of servers to host those domains. Preparing for a test usually takes a few years if you are just starting out, because you need to acquire domains that are several years old. At the same time, you can buy older domains to quickly acquire them. Just think about what you want to test, and start changing your domains to see what Google does. This is the best way to test the Google algorithm.

Article provided by Philip Rudy of ImageWorks Studio. Philip is the search and social manager there and has been in the Internet marketing space for the past three years.

Add Comment Register



Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>