To recap, around 4 months ago, on July 3rd, I moved the full blog from blog.outer-court.com to blogoscoped.com. With the move came a redesign, including some template amendments, but mostly CSS and image changes. There were also feature changes, and the archive index pages were relaunched.
The old domain had something around 9,000 - 10,000 pages indexed in Google when you checked with the “site” operator. The old domain also had a PageRank of 7, down from a previous brief spot in the PageRank 8 sunlight. (Google internally uses finer-grained numbers to express PR, so maybe Google Blogoscoped was somewhere on the upper edge of 7 at the time.)
Switching to the new domain, I put up a permanent redirect HTTP header for all pages. This worked relatively well with Google search results itself. Polling for site:blog.outer-court.com and site:blogoscoped.com first showed something along the lines of 9,000 to 0, then 4,500 to 4,500, then 0 to 9,000. The actual numbers weren’t as round but you clearly saw Google moving the pages one by one, and the overall balance was constant.
However, for many other tools accessing Google Blogoscoped the move had to be explained “manually,” as I previously described. The most trouble-some of all were moving iGoogle gadgets, so I finally just disabled the redirect for these gadgets (and perhaps Google will improve this process in the future).
Google Images on the other hand, again checked by using the site operator, shows a much, much slower update circle. Right now, the old domain has an image count of around 3,250 in Google, and the new one counts around 1,980. In other words, these numbers are still wrong after many months.
Today, I can see the Google Blogoscoped front-page has a PageRank of 5. That’s down by two points compared to the old domain’s PR7, but up 5 points considering blogoscoped.com itself – the domain existed before, but was unused – had a PR0. Three explanations come to mind to explain the new PR:
I think whatever the case, it’s safe to say that the new domain lost at least some amount of backlinks, simply because there are certain tools which don’t “get” a permanent redirect header. For instance, the technology behind the Technorati 100 listing wasn’t able to get this, and dropped Google Blogoscoped out of the listing pretty quickly after the move. That’s just one instance of many backlinks that might have gone extinct this way, causing PageRank-passing to go along with it.
Above you can see a graph of the traffic to Google Blogoscoped since April this year. The graph shows visits, and I marked the approximate position of the move in orange. As you can see, the traffic stayed pretty much the same. The average number of visits per day was 16,624 from January - June, and 16,043 from July to October, not counting the RSS full feed visits. (Note that this number of average visits changes depending on which months I use to compartmentalize the data – for instance, the number was up to 19,074 for a recent 31 days time frame.)
As this is not a “controlled” experiment – e.g. the quality and quantity of blog posts (or Google-related breaking news) might have gone up or down since July – constant traffic doesn’t necessarily mean that the domain move didn’t have any effect. Traffic is influenced by a huge number of things which are in constant flux, and it’s possible that two effects cancel each other out here, e.g. “Number of posts went down, but blog URL is easier to remember” or “New design appalls some visitors, but Google prefers stand-alone domains to sub-domains” etc. (to give just hypothetical examples).
Another number that is interesting are the referrers from Google. Whenever someone searches for something and clicks on this blog’s URL in Google, Google Analytics tracks it. As opposed to the overall traffic chart above, the following chart of the search referrers since April shows a slight drop at first, only for the traffic to pick up again later on:
As a conclusion, all in all, except for the PR drop, I think the domain move went rather smoothly.
>> More posts