This is not a bug with the counter, just a the limitations of counters period. The way the counter works is that when your browser requests the page of a website with the counter code, the code on the page gets triggered and sends a request to the Statcounter server for an image. This is the first hit. Once the site's server has delivered the page source to your browser, it no longer has anything to do except fulfill other requests. If the visitor's browser makes no other requests, like load another page or refresh the current page, no further hits are recorded to your project. The visitor might spend 10 minutes or more reading the page, or he may just have simply hit the back button. Their is no way to tell how long they were there. At least one more page load is required to make a calculation as to how long they were on the site. Think of it as someone ringing your door bell. You hear the first ring, and you wait,... and wait. No further rings. Did they leave? How long was the person at your door? Maybe they left after a few seconds, maybe they waited a few minutes. If the door bell rings again after 30 seconds you could make an assumption that they were there at least 30 seconds. Suppose that person rings once and then stands there for 10 minutes and leaves. - Visit time is given as zero. What if they rang once, then again after 10 minutes and left right away? - Visit time is given as 10 minutes. They were standing at the door the same length of time but the recorded results are different. You see how a second page load is necessary to make a calculation for length of visit. Even the server logs cannot tell you more about visit length. Certain web site layouts or content lends themselves to many zero second visits. A blog or page that is lengthy often has zero second visits. If you have a long article or story to read, break it into another page to help make a calculation.