Waxy.org is the sandbox of For my job, I recently researched and reviewed almost every major web log analyzer on the market. Almost every package provided the same basic level of detail, summaries of daily/monthly/weekly usage and aggregate statistics.
But we needed more detail. Much more detail. We wanted to track the paths of any user (authenticated or not) through the site, to see which pages they looked at, in which order, and for how long. We wanted to drill down to any time period, seeing who visited the site during that period and what they were looking for. Here's what I found...
Nearly all the commercial packages I reviewed had the same basic information as their popular open-source counterparts -- AWStats, Analog and Webalizer. The biggest differences were mostly skin-deep: ultra-fancy (but no more useful) 3-D graphs and pretty menus.
Take a look at the sample reports for WebTrends, Urchin (user/pass: english/english), HitBox Pro, 123LogAnalyzer, NetTracker, Wusage, Summary, and LiveStats.
As you can see, all of the above are fundamentally the same: static summary reports. Only two packages were able to provide the individual clickthroughs we need. The first, Funnel Web Analyzer, provides the bare minimum of clickthrough detail, doesn't run on Unix servers, and costs $1000.
The second choice was the best, by far: Flowerfire's Sawmill (see a sample report). It does everything we need, is extremely customizable, and also has the single-easiest Unix installation and configuration process I've ever seen. Custom log filters, dynamic filtering of data, static pre-generation of reports, and on and on. It's more powerful than any competitor by a factor of 10. It runs on Mac, Windows and Unix, and requires no additional software to be installed (including a web server or database).
Plus, it's cheap! Or free, if you help them test it for a couple hours. If you've written or know about a better log analyzer, I want to hear about it.
(Thanks to Leonard for originally mentioning it to me. Honorary mention goes to Clicktracks, which sports the most unique interface I've seen: a browser-like interface with statistics overlaid on the web page, perfect for usability testing.)
6:53 AM
Have you ever heard of TeaLeaf?
It's not really a log analyzer, and it costs a lot more than Sawmill or probably any of the others that you looked at, but I was completely impressed when our company had it demoed.
It captures all the information going from the webserver to the client, and allows you to replay sessions and view exactly what the user saw. It indexes everything on a seperate server (more expense), and has some extensive searching/filtering capabilities.
If you are building large-scale web applications/e-commerce sites. I would think it to be an invaluable debugging tool.
1:10 PM
And while we're at it; the 'Sawmill Pricing' page says it costs "$99 for an individual, $399 for a small organization". A similar pricing scheme goes for many other apps too (e.g. WS-FTP).
I've always had that question: how does the creator know the customer doesn't lie to him?
For example, a representative of a small organization may appear and claim he's bying Sawmill for himself. That way the company is saving $300.
I'm sure I'm missing something here..
PS: "Students can purchase Sawmill at a 75% discount (i.e. for 25% of the organization price, or US$99.75)" while it costs "$99 for an individual". Isn't that a bit weird? (A student paying more than an individual? I guess all students will try to hide their identity and claim they're just individuals...)
2:21 PM
I downloaded and played with Clicktracks, and it really was pretty cool. But $500?!? I was thinking about buying a copy until I saw the price. Why are all log analyzing programs so effing expensive? I think I'll just stick with Sawmill.
By the way, great post Andy. Thanks for the research.
8:48 PM
Konstantinos: I think the idea is that students can get a multi-user license for the price of a single-user license, if they want it. And to answer your other question, it's just the honor system. An ethical company will register the appropriate license, rather than risk getting caught by a disgruntled employee or an SPA audit.
10:51 AM
Sawmill is great, the biggest weakness (besides the interface / need to rebuild db's over and over while setting up) is that it's single threaded. Which is a mighty pain in the butt when dealing with big logs. Where are the MPI/PVM traffic analyzers is what I want to know.
[If you need hard core analysis and you can afford to pay a magnitude or two more and you don't mind shipping your data out for processing, Omniture's SiteCatalyst has some very slick features and a polished interface.]
5:10 AM
(I'm guessing) because I'm in Korea, the clicktacks site redirected me without asking to their Japanese site, with no way that I could find to get to English.
That was annoying. I'm just sayin'.
3:00 AM
Ha, I'm a little late in replying, but I haven't found one yet that suits my needs, so my husband made a program that shows who's on my site at the moment and what all they've viewed, complete with IP and precisely what time in Portugal. He made another that show what are the most viewed archives. We recently switched from our own server to one in the states that comes with an archaic version of AW and it severly sucks.