Over the last few months, I have been experimenting with Google Adsense and using the channels feature to measure the impact said changes.
There are volumes of information on tuning Google Adsense for higher performance on the internet, so I thought I might as well put some of their recommendations to the test for myself. As a small experiment, I have been been randomly selecting the colour of the Google Adsense ads on each impression for the last six months.
The Adsense control panel shows that over any given period of time, each colour or channel has been represented equally to the user. Throughout the rather lengthy experiment, I have gone through four different standard colour themes, which included:
- Open Air
The Google Adsense tuning guides suggest that blending the advertising into the site or using complementary colours yields the best outcome. For my personal site, which has always had a largely white based theme – I can confirm that a full blend has been the highest performer. The list of Adsense themes above is in ranked order based on the click through rates that they have generated. Surprising, at least to me, is that Open Air has not only beaten the competition each month, but at times by as much as 250%!
In the coming month, I’ll be looking to perform some testing on the size and placement of the Google Adsense ad blocks through the site. As soon as I have enough data to make a reasonably informed decision about the results, I’ll post them here for everyone else to see.
Rich Skrenta wrote an article recently about ranking web 2.0 sites by server performance, in which he talks about server response time and latency and how it impacts a site.
To see how everything stacked up, Rich decided that he’d profile over 500 of the top web 2.0 sites and throw in a healthy bunch of familiar faces as a yard stick. Some of the more familiar sites which were profiled were:
The average response times of the sites profiled varies wildly, ranging from a blazingly fast 6 milliseconds all the way up to a pathetic 15 seconds. It seems that for every 100 web sites you go down the list – it increases the average response time by approximately 75 milliseconds until you get to the outriders which skew the results.
Rich conveniently includes the web server used for the site if it was available, which as you’d expect features Apache and IIS heavily. What I found particularly interesting though, was to see where all of the super cool Ruby On Rails web sites sit within the list. You’ll notice that the programming language or platform isn’t specified within the list, so you’re probably wondering how I joined the dots – well it was the Mongrel web server which many Ruby On Rails web sites use.
Scanning down the list of web 2.0 sites, you might have noticed how many sites are running Mongrel:
- 1 – 100, three sites
- 101 – 200, two sites
- 201 – 300, six sites
- 301 – 400, four sites
- 401 – 500, seven sites
- 500+, two sites
The web 2.0 space has been dominated by people building out the next cool thing using Ruby On Rails, as it was the flavour of the month. Given that there are so few sites running Mongrel as a web site, either Rich happened to pick over 500 sites which generally don’t use Ruby On Rails or combining it with Mongrel isn’t the preferred mechanism anymore.
Everything else aside, the list does point out one really really significant thing; it doesn’t matter what web server or programming language your site or product is built in, poor design and architecture will lead to poor performance in nearly every instance. Apache delivering the fastest and slowest content within the list is evidence of this fact.