When the middle score is two crowds

The five-point courtesy question at a waiting room in Lat Phrao sat just above the middle when someone averaged it. The administrator was ready to call the month ordinary and move on to the next form. The cards were not ordinary. Nearly half rested on the top step and close to a third rested on the bottom step. The middle steps were almost empty.

An average hides a split room. We draw the five steps as five beads on a rail and count how many cards sit on each bead. If the beads gather at both ends, the sitting is about two experiences, not one mood. In that waiting room the high cards praised a particular nurse's habit of explaining a delay before people asked. The low cards were about the same delay with no explanation at all. The question had not failed. The mean had.

This pattern shows up on hotel breakfast cards and on parent questionnaires after a school open evening just as often as it shows up in a clinic corridor. People are rarely evenly spread across a polite scale. They bunch. When they bunch at both ends, a sentence like "satisfaction was moderate" sends the team home with the wrong job. One group wanted the explanation they already received. The other group never got it, often because it depended on who was on the desk.

Before you retire a question for looking flat, ask to see the steps. If you only hold a mean in a summary prepared elsewhere, ask for the count behind each step. A reading that starts from the mean and never returns to the cards will compliment a month that was, for a third of the room, unpleasant.

There is a practical consequence for the next form. If you already know a question splits, do not add more steps in the hope of nuance. Five was enough to show the split. Add instead a line that asks what happened just before the person marked the step: were they told the reason for the wait, or not. That line is what the sitting can act on. Another shade of "somewhat" is not.

All field notes · Write about a questionnaire