KEI is one of the most popular formulas used for keyword analysis. Here I re-examine and hopefully enhance the original.
Just as a disclaimer. I don't believe this sort of formula should be used religiously. They are just tools to sort data in a way that helps in keyword analysis.
Here is the most common version of the formula. In my case I use the Google Keyword Tool to acquire Monthly Global Searches (Searches) and Google's search engine to determine the number of Competition Pages (CompetitionPages).
The larger the KEI the more effective the Keyword should be in generating traffic. The basics of the formula are that the more searches there are the better. The less competition there is the better.
I analysed over 7,000 search phrases to see how the Searches/CompetitionPages relationship worked. The following graph shows all the results.
Colour coding has been added to show how the KEI formula distributes itself over the results. Red indicates the top 10% of the KEI values, green indicates the top 50% and blue the rest. The graph on the right is a zoomed in view of the core part of the graph on the left.
From this I noticed that the standard formula tends to spread very quickly to the right hand side, thus promoting very competitive phrases quite quickly.
Of interest. You can see horizontal lines of Search Scores, showing how the Google Keyword Tool has a limited set of values it returns.
Not all websites are equal. Some have more authority and can compete for phrases more than other. I believe the formula should adjust itself so that weaker websites focus more on low competition phrases while strong websites target high search volumes.
I decided to do this based on a websites PageRank. Here is the formula I came up with:
The constants were chosen so that a PR6 website would be equal to the standard KEI formula. As rank dropped, the relative effect of competition increases. I decided to make the PR0 website cut of its top 10% phrases around half a million pages. The following graph shows the new formula related to a PR0 website.
You will notice that the top KEI results are confined a lot more to the low competition area, confirming that this formula will promote low competition phrases more than the PR6 version.
Some phrases analysed my be less relevant to the target market. Other phrases may have a great conversion rate. I believe the formula should take this into account.
I've indicated it here as a simple multiplier. I personally give all my phrases a score out of ten. I then square the score to get my ConversionRate.
The KEI formula results vary from the very large to the very small. Some people have added a simple multiplier to make the numbers look more manageable.
I've decided to play with using Logs. This has pulled all my results into the region of +10 to -10
As I mainly use the formula for phrase sorting, the application of the log has no real effect. I need to think more about its effect on the general scoring. I'm also thinking about the KER formula and the way it controls scaling via logs.
I had to add the +1s to stop errors. I have already noticed that this can cause issues for very low counts. It's better to use something like 0.00000001!
To try and understand the data distribution a bit more I graphed the frequency that particular competition page counts occurred.
Interestingly it closely matched the equation shown.
I could not find any clear pattern for search counts.
Those two highly searched for dots up the top are "hotels" and "hotel". The most competitive phrase is "web site"!