In the last post, we presented the most and least commonly used menu items. We noted that a problem with analyzing aggregated data is the potential for outliers to skew our results. Today, to identify these outliers, we will move from looking at aggregate counts to examining how these counts are distributed.
The table below presents key statistics on the distribution of clicks for each menu item (mouse only). For example, it shows that the median user clicked “User Bookmark Item” 6 times during the course of the study, or equivalently, 50% of users clicked this menu item 6 times or less. Note: each distribution includes only those users who clicked the menu item at least once (if we included all users, even those that never clicked the menu item, many of the Q1 and median numbers would be 0, and the table would not be as informative).
As expected, many of the most commonly used items are heavily skewed with means much higher than the median. “User Bookmark Item”, “Back”, and “New tab” are three of the most heavily skewed menu items; for instance, the mean of “Back” is 10x the median.
Some outliers are more influential than others. For example, the max observation of “Bookmark this Page” makes up 28% of the total count for this menu item. Accounting for outliers and for the shape of the distribution helps present a more complete picture of the most and least commonly used items.
Now that we’ve discussed basic frequency counts and the distribution of these counts, we can move on to more interesting approaches to our questions. In the next post, we will examine the number of unique commands each user uses and determine whether menu interactions follow the 80/20 rule, where relatively few features account for nearly all the product interactions. In the future we will address the question of how long users spend exploring the menu bar before selecting each particular menu item.