To evaluate the Keyword Difficulty Checker performance, I used a simple correlation study to showcase if the difficulty prediction correlates to the actual ranking of each article.
Keyword difficulty and ranking distribution
Before we begin, I must point out some of the old articles on the site were produced in 2013 and only have 500 words with very poor keyword alignment and optimization. Because of this, they are completely excluded from this study.
As for the new set of keywords, in addition to the 30 articles that are already published, we have another 11 articles getting published as we speak. So overall, we have 41 keywords from old articles and 41 keywords from the new articles (all published or will be published in 2016).
Keyword difficulty distribution
Of the 82 keywords we targeted, 56 keywords (68%) of the keywords were ranked “Easy”, 18 keywords were ranked “Moderate”, and 8 keywords were ranked “Difficult” by the Keyword Difficulty Checker.
There are a few reasons why we targeted “Moderate” and “Difficult” keyword.
1) We wanted to test how Keyword Difficulty Checker accuracy
We wanted to have a list of keywords with different difficulty to see if the more difficult keyword would ranked lower.
2) Keyword research process still work in progress
While we knew “Easy” keywords would perform better, there were other factors we needed to consider, such as which buying funnel the keyword is in and how easy it would be for the writer to create an article. As a result we had to include some “Moderate” and “Difficult” keywords
Keyword Difficulty Checker Accuracy
We used the following criteria to grade how accurate the Keyword Difficulty Checker predicted the difficulty of the keywords.
Easy – the keyword ranks for position 1-30
Moderate – the keyword ranks for position 30-60
Difficulty – the keyword ranks for position 60 and higher
The keyword Difficulty Checker performed pretty well by predicting 70% of the time for “Easy” keywords, 94% of the time for “Moderate” keywords and 65% of the time for “Difficult” keywords.
This means 7 out of 10 keywords that the Keyword Difficulty Checker predicted Easy actually ranked within position 30 or higher.
Surprisingly, it predicted only 63% correctly for “Difficult” keywords. This means we do not want to discard all difficult keywords all together.
In fact, when we zoomed in and checked all 8 “Difficult” keywords, one of them was actually ranking on position 16 (page 2) on Google.
Correlation between the Keyword Difficulty Checker value and keyword ranking
When looking at correlation between keyword difficulty and keyword ranking, there is a positive correlation of 0.41.
In case you are not familiar with correlation, it looks at relationship of two values and how one influences the other.
For example, when looking at “weather temperature” and “how many winter jackets are sold”, we can predict there is a negative correlation. If the temperature is higher, less people will buy winter jacket.
In our case, when Keyword Difficulty Checker value is lower (= Easier), the keyword ranking did ended up lower (=1st position is the best, 100th position is the worst). Thus the 0.41 positive correlation value tells us that the Keyword Difficulty Checker does successfully predict easier keywords. Considering there are so many ranking factors, this result is pretty good.
The result we are seeing here is very exciting. It clearly shows not only the Keyword Difficulty Checker is working, but also our method of creating and optimizing content seem to be working as well. In fact, we only built some blog commenting backlinks and nothing else so the result we are seeing here is quite spectacular.
It is important to note that the data sample (number of keywords) we have is 82 so it would be interesting to test this out on a larger data set, which might show a better result.
Also keywords for different industries have different difficulties because some are more competitive than others, thus it would be interesting to test that out also.
It will be interesting to see how the ranking of these keywords get translated into traffic and revenue. I will provide that analysis in the coming weeks once we have more data for the new batch of articles.