Palletizing Throughput: Estimating Capacity Without Confusing It with Robot Speed

For a repetitive pallet sequence, estimate ideal capacity as 3,600 × cases per pallet ÷ seconds per pallet. The seconds must include every non-overlapping activity inside the cell boundary. Robot travel speed alone does not provide that time.

Then state how the ideal sequence relates to the scheduled period. If you apply a run-time factor, define its denominator and exclusions so that the same delay is not deducted twice.

Build a time budget

For identical picks, use:

T = P × t + e + s

Qideal = 3,600 × N ÷ T

Symbol	Meaning	Unit
N	Cases completed on one pallet	Cases/pallet
P	Robot picks needed for that pallet	Picks/pallet
t	Complete average pick cycle: acquire, move, place and return	Seconds/pick
e	Non-overlapping pallet exchange time	Seconds/pallet
s	Other non-overlapping recurring time, such as sheet placement	Seconds/pallet
T	Total recurring time for one completed pallet	Seconds/pallet
Qideal	Output while this defined sequence repeats without extra loss	Cases/hour

If pick times vary, sum the individual times instead of using one convenient fast cycle. If two tasks overlap, use the actual event sequence or a validated model; adding both full durations would overstate the elapsed time. Confirm the handling unit: MHI’s taxonomy distinguishes picking concepts from layer transfer, which is why a “cycle” needs an explicit definition.

Worked example: assumed inputs

Assume 60 cases per pallet, two cases on every pick, an eight-second complete pick cycle, and 30 seconds of pallet exchange that cannot overlap robot work. Assume no sheet handling. These are invented planning inputs for arithmetic, not manufacturer specifications or a measured project.

• Picks per pallet: 60 ÷ 2 = 30.
• Recurring time: 30 × 8 + 30 = 270 seconds.
• Ideal output: 3,600 × 60 ÷ 270 = 800 cases/hour.
• With an assumed 90% run-time factor: 800 × 0.90 = 720 cases/hour over the stated scheduled period.

Here the 90% factor covers additional lost time beyond the recurring sequence, such as interruptions not already included. It does not deduct pallet exchange again. We assume no rejected cases in this example; if rejected or reworked cases matter, define whether output counts all handled cases or only accepted cases.

Sensitivity: change one assumption

All rows retain 60 cases per pallet and the same assumed 90% run-time factor. Results are rounded to the nearest whole case per hour.

Scenario	Recurring seconds/pallet	Scheduled-period cases/hour
Base: two cases/pick, eight seconds, 30-second exchange	270	720
Exchange takes 60 seconds	300	648
Complete pick cycle takes ten seconds	330	589
Only one case is handled per eight-second pick	510	381

A change in group size may also change acquisition time or tooling. The last row isolates one variable to show sensitivity; it is not a prediction that a real single-case cell uses the same cycle.

Check the surrounding line

If upstream supplies fewer accepted cases than the cell can handle, the cell waits. If downstream cannot remove full pallets, accumulation eventually fills. A buffer can absorb a temporary mismatch; it cannot sustain a permanent excess arrival rate.

For example, an assumed arrival of 12 cases/minute against service of ten cases/minute for three minutes adds six cases to a queue, provided both rates hold and no other transfers intervene. Size and control a real buffer from arrival records and the complete sequence, not this example.

For the RFQ, provide product mix, cases per pallet, group sizes, observed arrival pattern, exchange method, shift calendar, and the exact definition of accepted output. The simulation guide helps evaluate interacting delays; the palletizing index covers the equipment decisions behind those inputs.