hmmm... there are a lot of blogs/questions about this subject, if the following explanation shouldn't help.
the time spent per visit is calculated as "total time"/"total visits". knowing that, there are two reasons why the total is not the average:
1) every day has most likely a different amount of users and shows the average of those visits. to get an average, you need to get the total time per day ("visits" * "time per visit"), sum it up for all days and divide it by total visits.
2) but, doing the above one, will double count all visits at midnight. if a visitors starts it visit before midnight and ends after midnight, it will count for both single days, but only once in the total header row. you can see the double counts if you add "visits" as metric and sum up the single rows.
to verify the number in the header row for "time spent" you need to sum up as written in 1), excluded by the double counts as mentioned in 2)
urs.boller is 100% correct here, calculating the time spent metric on your own is a lot more hassle than it's worth. There are also additional factors to take into consideration, such as the exclusion of single page visits (since there isn't a second hit to get the time spent).
The summary numbers at the top of each row in workspace apply as if the dimensions was a single value. For example, if you're looking at all of September broken out by day, the summaries at the top would be for the entire month of September, aggregated and deduplicated.
how do I excluded those midnight double visits ? I understand that the summary number in the column header exludes the midnight double visits since this summary number is smaller than the sum of each row .
For example, here, I can surmise that there were 2 visits counted double at midnight.
Visits Time Spent per Visit Day 8 x 1. 3 a 2. 4 b 3. 3 c
How do I get a total time as you explain in 1) excluding my double visits from each row ? When I finally divide by any total of visits ( with or without the double visits) I get a wrong Time per Visit header...