Traffic is picking up on the website grader and with the increased usage we are getting some great feedback from users. The SEO community is a tough crowd to please but when we started this project we knew that it would take several iterations to make something useful. If you are picking up on this project midstream, please read the background on building the web grader, the official launch and our first set of enhancements.
Last week we rolled out a very nice update to our score. The feedback actually came from a client who dismissed his good score saying "even though my score is higher, my competitor has more valuable keywords than me." In this particular case, the client's site had a forum that brought traffic but very few conversions. Their competitor, however, had a lower score but their rankings were on the keywords that tended to convert to leads and new business.
Adam and I talked about how to address this and the answer was something that we blogged about back in 2009... use the estimated cost per click to adjust the score. In this case we adjusted our formula from estimated traffic / rank to (estimated traffic / rank) * estimated cost per click. See the screenshot below:
Previously, the website in the left column had a score of 174 and the website in the right column had a 33. With the scoring adjustment the first website gets a boost because the first term's average cost per click is $6.75 while the rest of the keywords have an average cost per click of $.05. We believe the boost in the first website's score is correct given its ranking on a more valuable term.
Some people have asked if there is an easy way to infer traffic to a website from our score. We have steered away from those types of estimates because our experience from sites like Alexa and Compete is that they are not accurate compared to the real information reported by a tool like Google Analytics. Our score, rather, should be viewed as a way to quickly compare the relative success of SEO campaigns to a group of websites.
Keep the feedback coming and be sure to run some tests with your own data.