A data set of transaction times contains many values between 20 and 35 seconds and two rare system-failure times above 200 seconds. A product manager wants to describe how tightly clustered the typical transaction times are. Which measure and reason are best?