Fri, 04/09/2009 - 11:10 — kotoponus
I am coming from a commercial portal site background, and I believe that the web log analysis is the key to unlock the behaviours of users in your site. As you know, you can learn quite a lot of things: where they come from, how long users stayed, what contents they are interested in, movement of users, etc. etc. We can also data mine information to be used for more selective analysis as well.
So, I am relatively serious about how I deal with web log and I am generally inclined to say that you have a better control and knowledge of the data if you create your own analysis tool. However, one could be spending forever fine tuning the in-house tool as the data changes constantly and if you are low on resources, you may just be tempted the idea of using the publicly available tools. Which is my case.
Such tools are there to make your life a bit easier, but not a whole a lot easier. (With my limited experience, I have looked at AWStats, Webalizer and Analog. Wanted to look at Urchin, but it costs a dear.) Many of them look pretty, but the problem is that we do not exactly know what has been done to the logs when they spit out the results unless you dare looking into the source codes. They tend to have a certain level of configurable controls, but the basic philosophies of how they generate data varies and that is not so easy to discover unless you are willing to spend some time.
I have been also looking at the standard weblog feature which comes with standard Drupal install (primarily ver 5, less so for ver 6). Again, Drupal 5 has a report configuration, but generally speaking, I am not really thrilled by what I see. Again, compared to what I generated, the way they generate the hits is rather conservative in count. I am not trying say either mine or Drupal is better or worse here, but it is intriguing how the difference is seen like this. Again, the transparency of the generation logic is the issue here just like non-Drupal ones. The only nice thing, as far as I can see, is that Drupal seems to consolidate the aliases and node expressions. Certainly dealing with this could cause a headache as the log code needs to speak to the node table or whichever table that maps node and aliases. It is a very nice feature to have for Drupal specifically (and for some others which may work similarly to Drupal).
e.g.
node/5 = content/i-am-here;
1hit of node/5 and 1 hit of content/i-am-here;
It would generate 2 hit on "I am here."
Not that I am saying there should be an only way to deal with web logs, different circumstances call different solutions. But I have not found my way yet that I feel happy with, but what do I know I am relatively new to Drupal, and maybe there is a sound way that Drupal people use. Having said above, I am very curious to know if anyone has attained your enlightenment in this regard at least. ;)
Recent comments
2 days 18 hours ago
3 days 22 hours ago
6 days 18 hours ago
1 week 1 hour ago
1 week 4 hours ago
1 week 1 day ago
1 week 1 day ago
1 week 1 day ago
1 week 1 day ago
1 week 1 day ago