The assumption nobody writes down
Ask an operations team how they plan for peak and you will get a ratio. So many orders per picker, per hour, per shift. The ratio is rarely questioned because it has always been roughly right, and because the alternative, planning for a demand curve nobody can predict, sounds worse.
But the ratio hides a decision. It assumes the work arriving in an hour is the same kind of work that arrived in the last one. When that holds, headcount is a reasonable proxy for capacity. When it does not, and at peak it never does, the ratio quietly turns a capacity problem into a hiring problem.
The expensive part of peak is not the volume. It is the variance the volume arrives with.
Where the cost actually sits
Across the operations we have worked in, the marginal cost of a peak order is dominated by three things, and none of them are the picking itself.
- Rework, an exception found late, after the unit has already moved three times.
- Idle capacity, people staffed to a forecast the day did not follow.
- Escalation, a decision that waits for a supervisor who is holding forty other decisions.
All three are decision problems wearing a labour costume. Add people and you buy down the third one temporarily, at the price of making the second one worse. That is the linear link: not physics, but the shape of a system with no way to decide faster than a human can walk the floor.
Breaking the link
The operations that break the link do the same three things, in the same order, and the order matters more than the tooling.
- 01Instrument the work where it happens, not where it is reported. If the first record of an exception is a shift-end summary, every decision made that shift was made blind.
- 02Move the decision to the data. Routing, sequencing and exception handling encoded as policy, so the answer does not depend on who is on shift.
- 03Staff to the variance, not the mean. Once decisions are fast, a smaller core team absorbs a wider band of demand, which is the actual saving.
Note what is not on that list: replacing people. The teams that get the most out of this keep their experienced operators and stop spending them on triage. The floor knew what to do; it did not have a way to say so quickly enough.
What to measure next peak
If you take one number into the next cycle, make it the time between an exception occurring and a decision being made about it. It is unglamorous, nobody reports it today, and it predicts peak cost better than volume does.
