The Four-Day Warehouse Memo
Devon Park manages a 60-employee parts warehouse and wants to move everyone to four 10-hour days. Park's memo to the vice president rests on a six-week trial in the returns department: eight employees, four days a week, ten hours a day. Returns processed per hour rose 11 percent during the trial. Three facts are also in the memo, though Park does not discuss them. First, a new conveyor scanner went live in returns in week two of the trial. Second, two of the eight employees quit mid-trial and were not replaced. Third, unscheduled absences in returns fell from 34 hours during the six weeks before the trial to 9 hours during it.
Park's conclusion reads: “Ten-hour days make people more productive. Rolling the schedule out to all 60 employees will raise output by about 11 percent and cut overtime spending.” The vice president asks for a review before any decision.
1. Identify the two alternative explanations the memo fails to rule out, and explain why each one weakens the 11 percent figure.
Read the worked answer
1. The scanner. The 11 percent gain was measured in the same six weeks a new conveyor scanner went live. Scanning is exactly the kind of task that speeds up returns processing, so the gain may be equipment, not schedule. The timing is not a minor coincidence; it is a competing cause that arrived simultaneously with the treatment.
2. Attrition. Two of eight employees left mid-trial. The six who remained were not the same group as the eight who started. If the two who left were the slowest, the least able to tolerate ten-hour days, or the most resistant to the change, the remaining group's average would rise even if nobody's individual productivity improved. The 11 percent could be a smaller group doing the same work faster, or simply a different group being measured.
Either explanation alone is enough to block the memo's inference, because both produce a rise in output per hour without ten-hour days causing anything. Together they make the trial nearly uninterpretable as evidence about scheduling. A third, weaker concern: six weeks is short enough that novelty alone can lift performance.
2. The vice president wants one more piece of evidence before deciding. What should Park collect, and why is that a better test than simply running the same trial for another six weeks?
Read the worked answer
Park should run the four-day schedule in a department that did not receive new equipment, alongside a comparison department that stays on the five-day schedule, measuring the same output metric in both.
Why this beats extending the trial: a longer trial fixes only the novelty problem. It does nothing about the scanner, which will still have gone live at the same time, and nothing about attrition, which will keep thinning the group. Repeating a confounded design for twelve weeks produces twelve weeks of confounded data.
What the comparison adds: a control group estimates what would have happened without the schedule change. If returns processing in the comparison department stayed flat while the trial department rose, the scanner explanation weakens considerably. If both rose, the schedule was probably not the cause.
What the no-new-equipment condition adds: it removes the most obvious rival cause rather than trying to argue it away afterward. Park should also report headcount and turnover for the trial period, so the attrition effect can be checked instead of speculated about.
3. Rewrite Park's conclusion so that it matches the evidence actually available. Your version should still be useful to the vice president - it must recommend something.
Read the worked answer
A defensible revision:
“In a six-week trial in returns, output per hour rose 11 percent after the department moved to four 10-hour days. Because the same period included a new conveyor scanner and the departure of two employees, the gain cannot be attributed to the schedule alone. Absences in returns fell sharply, which is consistent with the schedule helping but does not establish it. I recommend extending the four-day schedule to one additional department without new equipment, with a matched department staying on five days, and reviewing output and turnover after eight weeks before any company-wide rollout.”
Why this version works: it keeps the 11 percent, because it was measured, but stops at “rose” rather than “caused.” It names the confounds instead of burying them, which protects Park's credibility when the vice president notices them anyway. It drops the unsupported projection to 60 employees and the overtime claim, which no data in the memo addresses. And it still ends with a recommendation, so it is a decision document rather than a disclaimer.