Five visits and one misleading headline
Take five invented visit durations: 10, 20, 20, 30 and 3,600 seconds. Their sum is 3,680 seconds. Divide by five and the mean is 736 seconds, or 12 minutes 16 seconds. The median is 20 seconds. Four of the five visits lasted half a minute or less.
Both calculations are correct. They answer different questions. The mean allocates the total duration equally across the observations. The median identifies the middle of the ordered set. Our arithmetic is an illustrative example, not traffic measured on this site.
Choose the summary after defining the question
NIST's statistical handbook explains that mean, median and mode are different measures of location. With a skewed distribution, the mean is pulled toward the longer tail. A histogram helps reveal whether a single central value is concealing distinct clusters or extreme observations. The handbook also notes that a mode for continuous measurements depends on how the data are grouped.
Sources: s1
Before calling it engagement
For these five values, a readable report can give the sample size, mean, median and full range, with the tiny list available underneath. It also needs a definition of 'duration': time between server requests, an active browser timer and elapsed tab-open time are different measurements. The hour-long observation might reflect a completed task, an idle tab or a measurement artifact. Reporting a larger average would not tell us which explanation fits.