A few years ago I did some consulting work with the Web and Information Services group at Brown University. The work consisted of test data analysis, information architecture recommendations, and usability feedback.
For the Physics Department, I analylzed their raw server logs, summarized the data, and provided recommendations such as the following:
- Ignoring the months with incomplete date (January and November), there was a higher level of requests during the spring than the fall. Page requests went up, however, in the fall. A difference of 5% of requests between the spring and fall is significant, though there isn't enough data for hypothesizing a reason. A difference of .4% for page requests is less significant. The months with the highest number of requests were April, May, and March. The months with the highest number of page requests were May, September, and March. What are visitors seeking during these months of traffic spikes? Should new content or links to content appear on the homepage during these months?
- Out of about 23 million requests, 1.6 million requests (about 7%) failed due to a server error. 7% is a high enough rate that you may want to monitor future server errors.
- More than 75% percent of requests were for MP3 files, with more than 65% of requests made for one specific MP3 file (/physics/userpages/students/yyyyyy/all/mlxq.mp3), with requests made for the specific file as recently as November. If the files are not legitimate, you may have a filesharing problem on your hands. If the files are legitimate, you should consider different hosting configurations to balance the load of the demand. Computing & Information Services can help with that.
After analyzing the data, I recommended a simplified site navigation and an improved information architecture. When the Web Services director presented my recommendations to the Physics Department, he reported that they quite happy with it and were planning to use it as their road map for building out their new site.
- Ignoring the months with incomplete date (January and November), there was a higher level of requests during the spring than the fall. Page requests went up, however, in the fall. A difference of 5% of requests between the spring and fall is significant, though there isn't enough data for hypothesizing a reason. A difference of .4% for page requests is less significant. The months with the highest number of requests were April, May, and March. The months with the highest number of page requests were May, September, and March. What are visitors seeking during these months of traffic spikes? Should new content or links to content appear on the homepage during these months?
- Out of about 23 million requests, 1.6 million requests (about 7%) failed due to a server error. 7% is a high enough rate that you may want to monitor future server errors.
- More than 75% percent of requests were for MP3 files, with more than 65% of requests made for one specific MP3 file (/physics/userpages/students/yyyyyy/all/mlxq.mp3), with requests made for the specific file as recently as November. If the files are not legitimate, you may have a filesharing problem on your hands. If the files are legitimate, you should consider different hosting configurations to balance the load of the demand. Computing & Information Services can help with that.
After analyzing the data, I recommended a simplified site navigation and an improved information architecture. When the Web Services director presented my recommendations to the Physics Department, he reported that they quite happy with it and were planning to use it as their road map for building out their new site.
I also analyzed usability testing data for the Engineering Department's website navigation. I looked at each test's data, made recommendations following usability best practices, and put together the following presentation (using the Web Services presentation template):
