This past June, Kate Fisher and I presented “Using Simulation to Determine a Retail Best Practice” at the International IIE Conference in Montreal, Quebec. The main objective of the presentation was to show how a simulation model can be utilized to take into account many aspects of a process, and help a retailer (or any company) determine a best practice for a core operational procedure. We simulated a very common retail process of stocking shelves in retail stores. The main focus of our model was to look at the sorting and handling of the cases of items via different manual methods and material handling procedures. It was very easy and intuitive to model the manual steps of stocking an item to a shelf: walking to a cart, moving a cart, obtaining a case, opening a case, placing the items onto the shelves, etc. Up to this point, it is simple and accurate to use engineered labor standards and some simple modelling in a spreadsheet to show labor times and the effect of job setup and process changes to the whole task. For many types of operations and tasks, this type of model will be as detailed as is needed.
However, in some of the industries we work in, there are additional layers of complexity in an operation. The complexity can be summarized into components for modelling purposes. Examples of these components would be worker search time to find a product on a shelf, additional time needed to move a cart or pallet around shoppers in the store, and customer service interactions that could take place at any time. These “fuzzy variable components” can be time studied, measured, and still accounted for in the spreadsheet model. However, they are usually modeled as an average value and spread or applied evenly over all instances in the model. This is where simulation takes your model to the next level.
A common example is a customer service time. An example is customer service times that range from quick interactions resolved with a few words up to many minutes spent attempting to solve an issue for a customer. Maybe customer service consists of an average of five to ten seconds per customer that needs help and involves answering a question or pointing them towards an item. However, what about the customer that is extremely upset and has decided they are going to vent to the worker trying to stock the shelves? This interaction could take many minutes, and involve the worker needing to find a manager or better knowledgeable employee to help resolve the issue. In this case, the worker might have been pulled off the stocking task for ten minutes. Simulation modelling has functionality that produces this type of randomness in the model. It utilizes statistical distributions determined from the data.
An example is: 300 customer service observations ranging from one second to ten minutes. The team would upload these values into the software. The software would determine a distribution that “best fits” the data. This distribution will then determine the customer service times in the model and vary them accordingly.
It is valuable to see how a larger customer service interaction affects the operation. How does it affect the progress of the worker, and other workers in the operation? Outputs and depictions such as these can show a retail management team how they should refine their expectations and projections when certain situations occur in the store.
While customer service time is the example illustrated here, there are many other “fuzzy variable components” that can be entered into simulation models to determine their total impact. Some other examples are: entering customers that are using a mobile app to shop, the selling of warranty plans for higher priced merchandise, customers not having the best experience with a self-checkout machine, and many more. In summary, using a detailed modelling tool such as simulation can help to model the impact of “Fuzzy Variable Components” of your operation.