Kwai Chung Fulfilment Centre: Satellite Hub or Slotting Redesign?
Golden Delta Logistics runs a single 18,000 m² fulfilment centre in Kwai Chung for a Hong Kong omni-channel retailer. The profile: 11,500 active SKUs, 7,800 orders a day averaging 3.2 lines, a 16:00 order cut-off, and 62% of order releases arriving after 13:00. Design productivity is 120 lines per operator-hour; actual is 96. Pick accuracy is 99.1% against a 99.8% target, and OTIF is 93.4%. The site runs two shifts on agency labour that makes up 40% of pickers, and annual picker turnover is 34%. Travel is estimated at 55% of pick time, and 280 slow-moving SKUs occupy the eight bays nearest the dispatch doors because they were slotted at start-up and never re-slotted. Separately, 28% of orders go to Tuen Mun and Yuen Long, and the landlord next door is offering 4,000 m² at HKD 22/m²/month. Management has already decided that "a second hub" is the answer and has asked you to endorse it.
1. Diagnose the productivity gap and decide between the Tuen Mun satellite hub and a slotting/picking redesign. Show the arithmetic that determines your order of priority.
Read the worked answer
Arithmetic first. 7,800 orders × 3.2 lines = 24,960 lines/day. At 96 lines per operator-hour that is 260 operator-hours; at the 120 design rate it is 208. The gap is 52 operator-hours a day, roughly 6.5 FTE on an 8-hour shift. Travel is 55% of pick time, about 143 operator-hours a day; a credible 30% cut in travel (ABC slotting plus zone/batch picking on a 3.2-line order profile) returns about 43 hours, or 5.4 FTE. That is most of the gap, with no new building, and it also attacks the 0.9% accuracy shortfall, because the slow movers currently sit in the golden bays and push pickers into congested aisles during the 13:00–16:00 release surge.
What the hub actually buys. It touches only 28% of volume. Splitting about 2,180 orders a day across two sites halves pick density per zone, duplicates safety stock and needs a second supervision, MHE and WMS configuration. Fixed cost: 4,000 m² × HKD 22 = HKD 88,000/month before racking, and realistically HKD 250,000–300,000/month all-in. Even an optimistic HKD 50 per order transport saving on 2,180 orders is HKD 109,000/month - it does not pay back on transport alone. Its real case is capacity and next-day reach into the New Territories, not productivity.
Decision. Sequence the redesign first: 12 weeks, reversible, low capital, addresses root cause (slotting drift, agency-labour density, late order release). Re-test the hub after six months against measured, not modelled, productivity and service data. Endorse capital only where the root cause is addressed and payback is under 18 months on measured data.
2. The retailer wants a single KPI set to govern whichever initiative proceeds. Propose three measures with targets, and explain for each how it resists gaming.
Read the worked answer
1. Lines per paid operator-hour, with a travel-minutes-per-line counter-metric. Target 108 within two quarters, moving to 120. Computed against engineered standards including an explicit travel allowance, so supervisors cannot hit the number by feeding easy zones to fast pickers or by holding agency staff off the clock. The counter-metric (sampled weekly from WMS task data) makes the travel assumption visible rather than buried.
2. Order cycle time, release-to-dispatch scan, reported as 95th percentile. Target 4.5 hours at P95, not the mean - the 16:00 cut-off with 62% of releases after 13:00 is exactly where OTIF leaks, and means hide it. Pair it with the share of orders released before 13:00, which forces the commercial function to own the release profile instead of the warehouse absorbing it.
3. Pick accuracy from a daily random audit of 200 orders, with error attribution. Target 99.8%. Attribution into slot-data error, pick error, pack error and labelling error stops improvement being claimed by reclassifying warehouse errors as carrier damage. Counter-metric: repeat-error rate by SKU and by picker cohort, which exposes slotting defects that a one-off audit would score as random.
Add cost per order line as the commercial anchor: productivity gains that require overtime or agency premium are not gains. Review the set monthly with one owner, and freeze the definitions for at least two quarters so trends are comparable.