Fable
A runner who got the menu number right but delivered the wrong dish
The grand restaurant in the east of the city added a serving window between the main hall and the back kitchen, with a runner standing at the window waiting for menus. His job was: guests order, he reads the menu number, passes it through the window into the back kitchen, then carries the dishes that come out to the corresponding table. To assess the runner, the manager installed a record board at the window that tracked only one thing—whether the menu number was passed to the back kitchen correctly.
In the first month, the record board showed a delivery accuracy rate of 97 percent. The shopkeeper praised the clerks at the regular meeting. But the number of customer complaints doubled. The complaints were concentrated in the serving step; the order-passing step actually had few problems—the clerks delivered sweet-and-sour fish to the table that had ordered steamed fish, took Table 3’s soup to Table 8, and once even failed to hold the plate steady, spilling half a dish onto the floor.
The shopkeeper laid the data from the record board and the complaint list side by side on the table and studied them all afternoon. The record board said the correct-delivery rate had risen from ninety percent to ninety-seven percent—the assistants were indeed getting better at picking menu numbers. The complaint list mentioned wrong tables, dropped plates and spilled soup—none of which had anything to do with menu numbers. The shopkeeper understood one thing: delivering the right menu is the first step, but the right menu is no guarantee of the right table.
The restaurant changed two things. First, it kept menu-number accuracy in the assessment table, but only as an intermediate metric. Second, it added a customer satisfaction tracker — after each table finished eating, guests checked satisfied or unsatisfied, which was compared against the menu number to calculate the proportion of guests who ended up satisfied after the menu was delivered correctly. That is the true metric.
The errand boy began to notice a pattern: the more complex the menu number, the more time he needed to verify it, and the higher his delivery accuracy became—because he looked a few seconds longer. But the ones he got wrong most often were familiar old menus—he glanced and handed them over, the number was right, but he did not notice that the old menu had changed its preparation last week, sweet-and-sour fish had a different side dish, and the customer did not recognize the dish that was served. The right tool was selected, but the parameter value was wrong.
The kitchen later split the metrics into three gates: the first gate looked at the accuracy of passing along menu numbers, the second gate looked at whether the dish matched the guest when the number was correct, and the third gate looked at guest satisfaction. Each gate had corresponding improvement measures—the first relied on menu training, the second on checking the dish-serving process, and the third on direct guest feedback.
A small line was later added to the sign: whether the food is served correctly is not only about the little window where orders are handed in. The empty plates on the guest's table are the real bill.