Sample Answer
Simulation of a Warehouse Order Fulfilment System
Introduction
Simulation modelling allows organisations to understand how their operational systems behave under pressure and how different decisions influence performance. Warehouse order fulfilment is a suitable system to model because it contains queues, human activity, resource constraints and time dependencies, making it ideal for Witness simulation. This report explains the chosen system, outlines the problems it faces, describes how the current model was developed and validated, and evaluates the most effective improvement strategy. The analysis is grounded in practical operational theory and the learning outcomes of supply chain modelling.
Description of the System and Problem Context
The warehouse belongs to a mid sized retail company that stores a range of packaged consumer goods. The order fulfilment system begins when customer orders arrive electronically. These orders enter a picking queue where operatives retrieve items from shelves. The picked items are then placed on a conveyor that leads to a packing station. After packing, orders go to a dispatch area where they wait for courier collection.
The system suffers from three main problems. The first is long picking time during peak hours because there are not enough pickers to meet demand. The second is congestion at the packing station because packed goods are not processed as quickly as they arrive. The third is variability in order arrival patterns that makes it difficult for managers to plan staffing levels. These issues directly affect customer lead times and create labour inefficiency.
The goal of this study is to model the existing system, validate it against real data and test potential changes that could improve throughput and reduce delays. The data obtained for modelling includes order arrival rates, average pick time, picker travel distances, packing time per order and conveyor speeds. The warehouse management system stores time stamped process data, which makes it reliable for simulation.
Modelling the Current System Using Witness
The first stage in Witness was to construct a process flow that mirrors the order fulfilment stages. Orders enter the system as entities. Arrival rates were based on six weeks of recorded data, showing an average of sixty orders per hour during quiet times and up to one hundred and forty during peak periods. Pickers were modelled as resources with an average pick time of seventy seconds per item. The conveyor was modelled with a steady speed and limited capacity. The packing station was represented as a single resource with an average cycle time of ninety seconds. Dispatch was modelled as a buffer that empties at scheduled courier collection times.
Verification was carried out by checking that the model logic matched the actual warehouse workflow. This involved stepping through the model in slow motion, confirming that pickers moved as expected, orders travelled along the conveyor correctly and station queues formed in realistic locations. Once the logic behaved correctly, validation was conducted by comparing simulated output with real data. Over a simulated day, the model produced an average throughput of one thousand and two hundred orders, which closely matched the real average of one thousand and two hundred and thirty. Queue lengths, resource utilisation and peak delays were also similar to recorded observations.
This validation increased confidence that the model realistically represented system behaviour. It also confirmed that the main bottleneck in the system was the packing station. Simulation showed that pickers were often waiting for space on the conveyor because the packing station could not clear orders quickly enough.
Problems Identified Through Simulation
The simulation clarified the underlying issues more precisely than on site observation alone. First, the packing station operated at nearly full utilisation even during non peak periods. This means the station had no room to absorb variation. Second, picker utilisation was uneven. During busy hours pickers were constantly active, but during quiet hours they were idle. This uneven pattern increased labour costs without adding throughput. Third, order arrival variability had a compounding effect on downstream congestion because even small surges created queues that took long periods to clear.
The model also revealed “hidden” losses. Conveyor blocking added small delays to almost every order, and over the course of a day these delays accumulated to a substantial amount of wasted time. The simulation helped quantify these losses, something difficult to achieve through manual observation.
Proposed Solutions and Modelling of Improvements
Several improvement ideas were considered, but simulation was used to test them before recommending changes. The first solution involved adding an extra packing station. When modelled, total throughput increased by twelve percent and queue lengths fell sharply. However, this option required additional labor and equipment cost, so further testing was needed.
The second solution involved redesigning the picking schedule. Instead of using a fixed number of pickers throughout the day, a demand based shift pattern was tested. During peak hours an extra picker was added, and during quiet periods one picker was removed. This reduced picker idle time by almost twenty percent, improving labour utilisation without reducing throughput.
The third solution focused on reducing variation. A batch release rule was modelled where orders were released from the arrival queue in controlled intervals rather than continuously. This smoothed the flow and reduced conveyor congestion. However, it slightly increased waiting time for some orders, which conflicted with the company’s aim to improve speed.
After comparing all options, the most effective solution combined two interventions. The first was the addition of a second packing station. The second was a revised picking schedule that aligned labour with hourly demand patterns. When modelled together, average throughput increased by nineteen percent and the longest queues were eliminated. Picker idle time dropped and overall lead time improved significantly.
The decision to choose these solutions was based on cost benefit reasoning. Although adding a packing station required investment, the increased throughput and reduced delays justified the cost. The updated picking schedule carried no additional expense but improved labour efficiency. Together these changes offered the strongest performance improvement and aligned with operational objectives.
