Google Analytics: Underreporting, or just reporting differently?

I gave a brief talk about Google Analytics to a bunch of OneWorld folks last week. An interesting question came up, about Google Analytics’ rumored underreporting of web stats.

There’s no question in my mind that GA does report less traffic than many other web analytics tools - particularly compared to offline tools that look directly at server logs. For instance, the Idealware site has been recently getting about 800-900 visits a day according to AWStats, one of the analytics packages that's offered by our webhost. According to GA, though, we top out at about 600 visits a day.

It's disconcerting, no question, but does that mean that GA is underreporting? Or is AWStats overreporting? I did some research on this topic, and came up with a number of factors that can cause the discrepancy. You decide what you think.

• People with Javascript or cookies turned off. Any analytics package that relies on a snippet of code added to the site will only report on visitors who have Javascript and cookies turned on. Folks estimate that this will miss between 11-15% percent of users.
• Robots and Spiders. Google Analytics fairly rigorously excludes automated processes (like the tools that search engines use to gather site information) from their stats. Other tools are likely to be much less rigorous, unless you're actually updating your tool by hand to account for new automated processes.
  
• Differences in visitor calculations. Any tool that doesn't rely on a chunk of code added to the site has no particularly accurate way to count visitors, as opposed to just arbitrary hits on a server, and is likely to over count visits based on things like people using proxy servers (coming in via AOL, for instance) and people who step away from their computer and then come back. Relying on cookies, as Google Analytics does, is much more accurate.
• Problems in adding the GA code. Some people say they are seeing enormous drops and inconsistencies. Other hypothesize, and I tend to agree, that it seems likely that they haven't effectively added the piece of GA code that it uses to track traffic.
  
• Google Iffy-ness. At the end of the day, there's also reports of just occasional weirdness. I don't think that it's likely that GA stats are just wrong most of the time. I do think, though, that occasionally there will be blips where it goes out, stops reporting for a short chunk of time, is slow to update, etc. It's a free tool, and it doesn't make sense to rely on it if you need 100% accurate stats at a moment's notice 100% of the time. For the rest of us mere mortals, though, it's likely fine.
  

Here's my primary sources for this (the comments, as much or more than the posts)

How reliable is Google Analytics?

Google Analytics is Under-Reporting, or is it?

Web Numbers: What's Real?