Domino’s is a well-known brand and is the number one pizza company in Australia and New Zealand. Across the Asia Pacific and Europe, Domino’s Pizza Enterprises (DPE) is the largest master franchisee of the Domino’s brand.
Problem
Domino’s has been growing very quickly and opening a lot of new stores. When opening a new store, a number of factors must be considered, for example, the right location, the layout of the store, and later labor scheduling.
So, there are a number of steps or activities that they need to do, including:
- Working with the operations team to determine the kind of products they offer and the processes involved in providing those items. Then modifying the model taking those factors into account.
- Spending a lot of time in the actual store under evaluation, watching and interacting with employees. If they are not able to travel, they use video footage.
- Using time and motion studies to determine the different preparation steps and their durations.
- Accessing a lot of historical data in order to understand the problem exactly.
- Engaging with different construction and equipment specialists.
All of this helps to make the model more accurate.
Solution
A digital twin environment needed to be established in order to test new concepts, store processes, and ideal layouts before they were implemented in the actual physical layouts of a store.
Simulation to improve store layout
The first step when starting modeling is taking an actual store floor plan and converting it into an equivalent AnyLogic representation.
Domino’s identified three components to focus on in order to develop a simulation model to improve their stores:
- Physical store layout – the dimensions and shape of the floor space as well as the location of the store equipment. This needs to be converted into an AnyLogic representation and then run as a simulation.
- Process flows – this is essentially all the underlying logic that determines how the agents will act in the simulation environment.
- Order and staffing data – all the information on orders and personnel, as well as their roles that aid in delivering those orders.
The focus is on modeling for peak times and how to improve store efficiency during those times. The objectives are to design a layout that works best for any given store and to increase that store’s efficiency in relation to the layout.
An initial and important question was which floor space works best – square or rectangular? Through simulation it was discovered that square functions better because it allows for more design possibilities and more range of movement for the staff. A rectangular floor, on the other hand, requires more labor, higher order times, and results in lower customer satisfaction. Therefore, the idea was to focus on more compact and proportional floor spaces.
The next step was to optimize the layout of the store. The goal was to ensure that all staff members are moving in a more streamlined way. In essence, the objective is to minimize the number of steps necessary to complete an order while ensuring that the workers don’t obstruct one another.
In the illustration above on the left, the driver and makeline staff intersect and get in each other’s way and during peak times this can significantly reduce efficiency. The arrows illustrate the routes taken and they are unnecessarily long.
The alternate arrangement on the right was developed using AnyLogic simulation with no intersection of driver and makeline staff. The number of steps of both were reduced and the product could be delivered much quicker. This is an objective in all stores.
Some of the designs that have been created through simulation have already been turned into actual stores as can be seen below. This illustration shows a store in Australia that was damaged by flooding and needed to be rebuilt. A simulation was then created, a store plan developed, and now the store is currently being built.
Simulation for staff scheduling
A store, of course, also needs staff and determining the best labor scheduling for a particular design layout was the last task. Both understaffing and overstaffing have their own particular problems, but through AnyLogic simulation, the right number of staff on a given shift, for a particular layout, could be determined.
In one instance, at a new store in Australia, the builders came up with one layout (the baseline) and the model developers created two alternative plans. These designs were compared to show the efficiency. The outcomes of this analysis are shown in the illustration below.
Adding drivers reduces delivery time, but there comes a point when doing so offers no further advantage. The ideal spot is between 6 and 7 drivers for all 3 layouts. The decision maker can then select which layout they want to go with and where they wish to operate on this staffing curve.
A key insight from this store simulation was that Domino’s was able to identify efficiencies and reduce dispatch times for their products by 4.5%.
Results
Some of the results have already been illustrated above, but in summary, the following can be gained from this case study:
- The makeline, oven, and cold room should be arranged close together in a triangular formation to allow the makeline staff to easily move around efficiently.
- The production of food in front of customers, known as "food theatre," may be the root cause of sub-optimal layouts.
- Smaller, more compact floor plans can overcome poor design.
- For better layouts, proportional floor space is paramount (square is better than rectangular).
What-if analysis is continuously being worked on and is taking on a lot more significance. What would happen if a process changed, and would the store layout, for instance, need to alter to accommodate such changes?
The case study was presented by David Federer and Dr. Shelvin Chand, Domino’s Pizza, at the AnyLogic Conference 2022.
The slides are available as a PDF.
