If you have access to a server access log statistical analyzer app (like Webalizer, Analog Stats, or Awstats) on your server (typically via a server control panel like cPanel), then you may see something like Unique Visitors in your stats. Unique Visitors can be VERY misleading, and not very accurate.
Why? It does not take into account dynamic IP addresses. A dynamic IP address changes each time you boot your machine or network device. It also changes when you dial into your Internet service provider. Some ISPs change your IP each time you log on to their network while some make the changes periodically. Static IP addresses, however, do not change routinely like dynamic IP addresses. If a user is using a dynamic IP address, and the server defines uniqueness on the basis of IP address, the same person could be counted as multiple uniques. This is very common, and as a result, renders server log-based unique stats very unreliable and inaccurate.
Web-based stats like Google Analytics get around this by using other parameters including cookies or sessions to fine-tune uniqueness testing. The latter (cookies and sessions) are used to track users in order to determine if they have visited a web page in the past. This is dependent on the life of the cookie (how long the cookie persists before it expires), and if the user has cleared their browser cookies which is common. So, these metrics are also flawed in terms of uniques.
So, in a nutshell, the best approach is to use a number of methods (server-based and web-based) to collect and analyze traffic stats. And then compare, contrast and weigh the results to derive a guessestimate of what the “real” results are – especially as they relate to unique visitors. Sometimes, knowing the relative numbers are more powerful than knowing the absolute numbers. There may be a high degree of uncertainty in the absolute uniqueness values, but there is value knowing the relative numbers. For example, how the numbers change over time – especially in response to a new marketing campaign. Like all metrics, they need to be taken with a grain of salt – and in some cases, a dump truck load of salt.