Jim Gianoglio over at LunaMetrics posted an interesting blog yesterday, suggesting that there is value in tracking the number of keywords referencing your site over time. He laid out an interesting argument that it is supposed to give you a good indication of how well you fair on search indexing and whether there are issues with your site design. He ended with a request for someone to automate the effort for him, since his solution involved manually copying numbers month by month from the Google Analytics interface. Well, hey – that’s what we do with Next Analytics – automated web analytics!

The Approach

Since there was some question about whether the number of keywords simply tracks the number of visits, I thought I would expand on his report and include the number of visits as well. Jim also mentioned branded versus unbranded keywords, and I thought it would be good to see the difference. Finally, I have always found keywords to be a bit of a challenge because most of them are unique queries, so I thought it would be nice to see how many were actually used more than once (a.k.a. ‘popular’). This is all easily accomplished by downloading a list of keywords and visit counts, plus a handful of Next Analytics script commands.

Add One Month at a Time

With the sheer volume of keywords that some sites have (25 thousand is not uncommon), if you want more than a handful of months, a single data query would take a while and involve a lot of data, so I opted to process one month at a time and append it to a running list. That involved using an Excel formula to figure out what date was last queried, and from that, the new query range. The result is kind of slick – click the Refresh button and a new month is appended to the data, so you enter a starting date and add months one click at a time. On a monthly basis, just Refresh to get the latest info added in. This same technique could be used for any periodic analysis report.

The Results

Well, I was a little disappointed by what I found. The numbers actually do track the number of visits pretty well. I even tried the report on 5 very different web sites that have gone through some significant change over time. Still no wonderful AHA! moments. What was interesting was the difference between total and unbranded keywords; a couple of sites had almost no branded queries. Two others tracked by a noticeable, but steady, margin. Only one had significant variation, but it was based on very low volumes. All in all, it was worth finding the ratio between branded and unbranded, but the trends provided little insight.

If you want to try the technique for your own month-to-month metrics, download the workbook and Refresh it with Next Analytics. Although I displayed the results with the most recent date to the left, I have included notes in the comments if you prefer it the other way around.