Two Supervisors Read the Same Pilot Data
A warehouse ran a twelve-week pilot of four 10-hour days instead of five 8-hour days for its 40 order pickers. Results reported to supervisors:
- Lines picked per hour: 210 before the pilot, 214 during the pilot.
- Overtime hours: up 18% during the pilot.
- Order error rate: 1.2% before the pilot, 1.9% during the pilot.
- Sick days: 31 before the pilot, 27 during the pilot.
Supervisor Maya says, "Productivity went up, so the four-day week is working and we should keep it." Supervisor Owen says, "Errors got worse because people are exhausted by the longer shifts." No picker was asked about fatigue, and the error rate was not broken down by day of the week.
1. Which parts of Maya's statement are supported by the reported figures, and which are not?
Read the worked answer
Supported: lines picked per hour did rise, from 210 to 214 - about a 1.9% increase (4 ÷ 210).
Not supported: "the four-day week is working" and "we should keep it" go beyond the data in three ways.
- Per-hour output is not total output. A 10-hour day contains more hours, and overtime rose 18%, so the gain per hour may have been bought with extra paid hours. The report gives no total-cost or total-output figure.
- Quality moved the wrong way. Errors rose from 1.2% to 1.9% - an increase of 0.7 percentage points, which is roughly a 58% relative increase (0.7 ÷ 1.2). Rework and returns are real costs that the productivity figure ignores.
- "Should keep it" is a judgment, not a measurement. The data describe what happened; they do not weigh the speed gain against the error and overtime increases.
A defensible version of Maya's claim: "Output per hour improved slightly, but error rates and overtime also rose, so the pilot's overall value is unclear."
2. Owen says errors rose "because people are exhausted." What is missing from the report that would be needed to support that explanation, and what is a competing explanation the same numbers allow?
Read the worked answer
What is missing:
- Any measure of fatigue - a survey, or the error rate by hour of shift (errors in hour 9 and 10 versus hours 1–4).
- Whether the same people worked the same tasks; a change in order mix or in new hires could raise errors.
- A comparison group still on the five-day schedule, so ordinary seasonal swings could be ruled out.
A competing explanation the same numbers allow: longer shifts mean more hours worked per day, and error rates typically climb with consecutive hours on task regardless of schedule. Under that reading the four-day week is not the cause; the 10-hour day is, and a different arrangement (for example, staggered start times) might capture the per-hour gain without the error increase.
Key point: the data show two things changing together. They do not show that one caused the other, and no picker was asked how tired they were.
3. Owen also claims the sick-day drop "proves people are healthier on the new schedule." Evaluate that reasoning.
Read the worked answer
The reasoning is weak, though the direction is favorable.
- Four days is a small number. 31 to 27 is a drop of 4 sick days across 40 pickers over twelve weeks. A swing of four absences can easily be chance or a mild flu season.
- Schedule can change reporting, not health. With a four-day week, a picker may schedule an appointment on the off day instead of calling in sick. The absence disappears from the count even though nothing about health changed.
- "Proves" is the wrong word for any before-and-after comparison with no control group. "Is consistent with" is the honest phrasing.
What would strengthen it: compare sick days per scheduled workday, not raw totals, and compare against other departments that stayed on the five-day schedule over the same twelve weeks.