Oh, to sashay down a catwalk affecting an attitude of supercilious cool as the paparazzi gape in hushed reverence at your warehouse, which is absolutely comme il faut.
The odds of this happening may be slim, but it’s perfectly possible, and reasonably easy, to design a facility that could, in theory at least, become the talk of the town. Ikea, the Sweden-based retailer of furniture and home accessories, used simulation to reconfigure its complex distribution centers to support direct in-store, mail order, and online sales. The modeling technique, developed by Ikea and Establish, Inc., a Fort Lee, NJ-based supply chain consulting firm, not only produced facilities that delivered products to customers in the most efficient way but also became an operational tool that could be used years afterward.
Until the simulation model came along, Ikea was handling its direct-to-customer orders by picking the goods on an aggregated level in the central warehouse and shipping them to different local centers, where the items were later sorted and distributed to customers. It became evident that this method was not optimal in terms of cost and service level; furthermore, an increase in online sales raised fulfillment costs.
Ikea needed a way to evaluate and compare different distribution strategies. To obtain such a tool, the company decided to create simulation models that could produce data to serve as guidelines and support in the decision-making process. The model was originally designed as an analytical tool for a distribution center in Torsvik, Sweden, and is flexible enough to be used in multiple scenarios.
Baste and snip
The most important event in developing simulation models is creating a well-defined objective. In Ikea’s case, the main question was “To what degree do the different systems succeed in reaching the service goal, and what is the total logistics cost within the defined system?”
This preliminary phase of questioning should set parameters for simulating different parts of the operation. Furthermore, it should describe processes and data requirements in a structured way; for instance, by using integrated definition (IDEF) methodology. IDEF graphs that showed all relevant processes at the decided level of detail were approved by Ikea as blueprints for the simulation model.
To understand Ikea’s operations, the Establish consultants visited and assessed the company’s facilities and interviewed managers at Ikea headquarters. Ikea representatives were asked to come up with a number of key questions that the simulation model had to answer.
Based on the visits and interviews, three sets of IDEF charts were produced. Each described the process hierarchically, going from the general to the specific detail chosen for each activity. The charts included control parameters and resources. Written descriptions for the concepts were also created to ensure that everything was fully understood and agreed upon. The level of detail in a simulation model is important: If the level is too low, you spend too much time on trivial information; if it is too high, you do not get the results you need. Establish presented the IDEF charts and facts collected to Ikea during a one-day meeting at which the data were adjusted until everybody involved was satisfied with the content.
The Ikea project is similar to other simulation projects in terms of its steps and methodology. Briefly, the project covered the following processes:
Preliminary phase
- Research — interviews with key personnel and visits to warehouses
- Process analysis by using IDEF
- Client’s input on IDEF specifications
- Data requirements defined
Model building
- Start-up session and brainstorming
- Start-up input data collection phase
- Prototype — data shells 1 and 2
- Model building — version 1
- Output review 1
- Model building — version 2
- Output review 2
- Input data collection phase finished
- Model building — final
- Model acceptance
Debugging and validation
-
Education of client staff to run model and make analysis
-
Handover to client
On the runway
The Ikea model is non-deterministic — that is, the system has to be run for a “warm-up” period to establish a balance in the model. The simulation does not have a defined start or stop cycle, but is run continuously for a month.
The simulation starts at the picking position and ends as the item is set for delivery at the loading dock. To get the right level of detail, some processes are left out — for example, buffer storage operations and some aspects of distribution. The diagrams below and opposite show the model’s interacting parts.
A typical warehouse model is divided into zones so that articles can be directed to locations depending on how the item list is configured. The Ikea zones are as follows:
- Storage area for goods for manual picking and automatic sorting.
- Storage area for goods for manual picking and manual sorting.
- Crane storage.
- Picking position for crane goods to auto sorting.
- Picking position for crane goods to manual sorting.
- Sorting machine.
- Postal handling of small orders.
- DDC goods and customized products such as sofas and countertops.
- Merging position for manually picked products.
- Area where complete orders are merged depending on destinations.
The user controls the distance between different zones and functions. The distance between locations depends on aisle widths. Operators pick according to pick tickets. Manual sorting is the default, but an automated sorting machine could be activated. Carriers collect goods for various destinations according to a schedule.
The simulation handles each article number and each line in the picking order separately. It determines which item to pick first and where to pick it, selects the picking equipment needed, and assigns it. It then picks the item, moves it, and lets it go, waiting in a queue if necessary. Conveyors and other equipment merge the item with others in the order, until the item reaches the loading dock. Among the factors the model analyzes are processing time, resource utilization, and queue times and lengths for the various functions. Working schedules for different resources are applied in the logic. Sorters and other stationary equipment are simulated at a higher level; sub-functions and detailed parts of the machines are not simulated.
Animation is created in a two-dimensional, top-view version that shows individual resources and staff. Movement by staff and handling equipment is shown on a generic level. Complete customer orders from each picking area, and later the merged orders, are displayed as one entity in the model.
Prêt-à-porter
Ikea and Establish needed to specify what output the models should produce to be able to decide the input needed. Cost, equipment, scheduling, and shipment information was collected and put into the model as shown at left. The simulation reads historical data from actual Ikea order files. Varying the number of orders and items on each order row ensures the right combination of items in the order. For example, when a bed is ordered, the model factors in the percentage of times that legs are also ordered.
Employing a variety of parameters, the analysts were able to set up a user interface that could be applied to different scenarios. Stochastic variables were added to the model, based on statistical distributions built from historical data.
Most of the output data was presented by confidence intervals. All data were exported to Microsoft Excel or Microsoft Access programs for processing and presentation. The simulation model was built in ARENA, with input and output done through Excel spreadsheets.
At the completion of the study, the model was handed over to Ikea. The company uses it today for strategic analysis, product replacement, growth scenarios, and scheduling.
Ikea uses the results from the analysis of the model’s output as a foundation for decisions made when investing in new distribution centers. By applying the model, Ikea is able to pick the strategy that delivers the best service level and cost for online DTC sales. Adding growth scenarios enables the retailer to anticipate challenges and outcomes because the impact has already been tested before the first brick is laid.
In addition, simulation provided Ikea — as it does all companies — a deeper knowledge of all the components and requirements that went into the model and the processes and systems that will result. A modeling project often gives new insights even before the simulation is run because of the exhaustive data collection and analysis that have been done.
Simulation does not help companies with marketing and sales activities, but once the product is sold, it increases customer retention and satisfaction by getting the right product to the customer — on time and at the right cost.
Jeff Behn is a vice president at Establish, Inc., a supply chain consulting firm based in Fort Lee, NJ. He can be reached by phone at (201) 302-5930 and by e-mail at [email protected].