I recently encountered a situation where a client was confused by a phenomenon that, at first glance, looks very odd indeed. They had set up a report based on a single page, meaning that they excluded everything but a particular page from their query, and were returning visits. When they, by chance, looked at unique visitors, they were shocked to find that this number was larger than their number of visits. Let's have a look at this and see what sort of explanation we can find for this bit of strangeness.
Note: if at any time you see a term that you don't quite understand, try having a look at my Web Analytics Primer.
Visits are counted on the first 'hit' or pageview for a given session and can encompass a number of unique pages and pageviews. Visits are counted on the 'session' level, and so cannot be related directly or seamlessly with stats that describe individual pages or uniquely identifiable visitors.
Unique visitors is a raw count of visitors. There is, actually, no such thing as "Non-Unique Visitors", but "unique visitors" is the term used in Google Analytics, so we'll have to live with it. Unique visitors are counted on the 'visitor' level, rather than per page or per session. There can be multiple visitors for a given page (regardless of where or how many times that page occurs across all of an unique visitor's sessions), and a visitor can have multiple sessions.
Looking back at the introduction, it appears that the situation arose because we decided to include only the visits and visitors corresponding to a single page, and filtered out the rest. When we remove this filter, the count for visitors and visits reverts to something that makes more sense - considering all the pages on a website together, visits should always be greater than unique visitors. Because this is the norm, we expect it to apply in any case, even when we filter out everything but one page. However, things start looking differently sometimes, and it's because of the different ways in which visits and visitors are counted.
Since visits are counted just on the first pageview and are counted on the "session" level, when we try to separate one page out of it all we might find that we have fewer visits than visitors for that given page.
In this case, when looking at a single page on its own, what "visits" really means in this case is "visits where the particular page we're looking at was the first one encountered during that visit". Visitors, however, being on the "visitor" level, still means what it always means: "any visitor who ever visited that particular page or group of pages". This is why you might see a very small difference in visits and entrances (usually single-digit percentages).
Note that this differs subtly from "entrances" because multiple entrances can occur and still be counted as a single visit, as long as both entrances occur within 30 minutes of one another.
Let's have a look at the two visits shown above. Both visitors have two visits where they viewed three pages each. The first visitor hit PG8 as their first page for one of their visits (in green), while the second visitor hit PG8 as their second page on their visit (in red).
The first visitor's visit, or session, begins on PG8, and they get counted - we'll see them as a visit when we look at just PG8 (that's why it's green - it got tracked).
The second visitor's visit begins with another page, PG5, but they view PG8 later. When we look at PG8 alone, that particular visit is not counted (which is why it's red - it didn't get tracked).
However, both of these visitors, at some point, viewed PG8, and so both are counted as visitors who have been to PG8 when we look at PG8 on it's own. For these two visitors and PG8, we will have counted 2 visitors, 1 visit, and 2 pageviews.
This all just goes to show you that things are not always as they seem when you start playing around with filters. A general principle to follow is that levels adjacent to one another in the "visitor, visit/session, page" hierarchy can influence the level adjacent to it. Consider the visitor and page to be adjacent, as though each level were one of the points on a triangle, as you can see in the image above. Now to seek out another mystery to solve!
If you have an unsolved mystery, be sure to leave a comment.