There are two major types of analytics systems: client-side and server side. Client-side analytics relies on an event fired by a user to record a page-view. Google Analytics is the most common client-side system. Client-side packages are beneficial because they can also track non-page-loading events, such as interacting with a form or video. Being user-ran, they can also harvest user data like screen resolution and connection performance. A server-side system, like AWStats, looks at server logs to determine the volume of pages requested. Server-side analytics are good for tracking special cases– like lost pages that need redirects, the traffic of search engine spiders, and mobile users, but have limited insight on conventional PC users.

It’s important to recognize that client-side and server-side analytics never match exactly. Since a client-side system cannot record traffic by robots, and some limited users– like mobile users with no image or script support, it tends to undercount by a few percent. Server-side systems often mis-classify users based on browser headers- many obscure browsers emulate IE or Safari.

Moreover, even inside a category, disputes occur. Does a visitor who sees a second tab on one page count as a bounce? Is a user who hits a Bing ad driven by Bing or the Microsoft ad network? Analytics vendors have many judgement calls. An important guideline is to use analytics data for month-to-month comparison in a single vendor. Google Analytics for June compares sensibly with Analytics for May, but trying to reference it against AWStats for June leads to confusion and bad decisions. Occasionally, a disparity between packages can reveal unusual user behaviour, such as a denial-of-service attack seen by the server-side system and not Google Analytics, but it’s more often statistical noise.