Inventory optimization refers to reducing excess inventory, avoiding lost sales due to unavailability of items in stock, well-defined safety stocks and reorder points, and much more. In this article I demonstrate how simulation can be used to master the challenges of inventory optimization. A commercial project example from a pharmaceutical supply chain (case study) underlines my statements and analysis.
Real-world inventory optimization challenges
Consider this: What happens in our customers’ warehouses?
- “I have too much inventory”
- “I lose sales because I don’t have the right inventory”
- “There is not enough safety stock”
- “Reorder points were defined a long time ago”
- “We have always worked with those lot sizes, why should we change them?”
- “The warehouse has too much obsolete inventory and causes too much scrapping”
- “The delivery times of my supplier are very variable and affect my planning”
If you are a warehouse, supply chain or operations manager then some of the these statements sound familiar to you.
Inventory management and planning encompasses a wide range of operational complexities dictated by internal restrictions and particularities of a business. Moreover, inventory management is also dictated by external factors caused mainly by variations in supply and demand throughout the supply chain. Traditionally, inventory policies have been defined by analytical models. Policies used to be rigid and built on premises that do not adjust to the ups and downs of todays’ supply chains.
Analytical models and static formulas fail to capture the dynamics of inventory planning. In addition, they do not allow for safety stock or inventory policy validation before their actual implementation and deployment. A trial-and-error approach, however, can cause downtimes before a “good-enough” solution has been found.
So, what is the correct approach to this problem?
Simulation considers dynamic inventory developments
A simulation model describes a system and the rules of action under which it operates. This includes relevant interdependencies between the various system elements as well as the logic of related business processes. Furthermore, a simulation model allows planners to test and experiment in a risk-free environment, while capturing all the details and complexities of a supply chain. Such complexities are caused by e.g. random or time-dependent effects, and effects that interact within the system, such as fluctuations in demand, variability in delivery times, changes in lot sizes or order frequencies. Many of these effects are not considered by analytical models.
And there are more advantages offered by a dynamic simulation:
- Performance monitoring over time
- No black box model
- Spotlight on relevant trade-offs
Performance monitoring over time
A simulation study models inventory levels over time. In that way analysts can examine incremental changes and spot breakpoints in planning. An exemplary simulated inventory projection time-series can be seen in below figure.
Requirements exceeding available inventory result in backlog. The red curve in above chart shows backlog development. It represents unavailability of the requested product at required shipment date. This has a negative effect on sales revenue and customer satisfaction.
Inventory simulation is not a black box
Another advantage of inventory simulation is that the user can see why the optimal stock policy has been selected. Moreover, visualizing inventory simulation runs helps users understand how inventory policies have been defined. Unlike analytical models, no assumptions or abstractions are needed.
Inventory simulation puts trade-offs under the spotlight
A simulation model facilitates visualization and evaluation of trade-offs between customer service levels and inventory costs and, therefore, allows for defining the optimal inventory investment allocation. I illustrated this in below figure.
Above chart compares the current vs. various optimal policies. All optimal policies are pareto-optimal. That is, no improvement in one of the two performance indicators (out-of-stock ratio and investment sum) can be achieved without a detoriation in the respective other performance indicator. The results are derived from simulation runs using a inventory simulation model.
Simulation allows for scenario and constraint analysis
A simulation study can test inventory operations over a wide range of scenarios with widely differing constraints (area, volume, weight, working capital, capacity investment, etc.). Additionally, new planning parameters and stochastic processes can be included by the inventory simulation. Based on these considerations, business decisions can be made based on “near” real-world operating conditions and constraints.
Exemplary problems that can be solved with simulation
Based on my experience I list several challenges in inventory planning that can be solved with simulation:
- Inventory policy optimization: I challenge the current policy and define the best set of planning parameters for each SKU with management indicators specific to your business and service level goals. The new policy will not be arbitrary and will take into account variables such as lead times, lot sizes, review periods and costs.
- Safety stock definition: Based on fluctuations in demand, variability of delivery times, operational interruptions, and industry specific risks I can define appropriate safety stock levels for each product of your business.
- Quantification of better forecasts: More robust forecasts have a great impact throughout the entire supply chain. Improved forecasts reduce operational uncertainty and, therefore, required safety stock. A simulation-based approach can quantify the effect of forecast accuracy on inventory costs and service levels.
Commercial inventory optimization case study
A pharmacy chain in Latin America was suffering from stockouts and backorders for several products, as well as large stockpiles of other SKUs. This problem occurred in all of its distribution centers. Inventory policies defined largely by intuition and highly variable demand led to poor operating performance and excessive capital costs.
Based on these restrictions and challenges, we worked together with the company and defined an optimal inventory policy for its finished products. This allowed the pharmacy chain to achieve the targeted out-of-stock levels while at the same time being able to reduce working capital.
After modeling the company’s MRP with its inventory policies and simulating distribution operations with different planning parameters, we developed several scenarios that reduced inventory costs and stock outages. The selected scenario generated savings of $0.8M in inventory investments and decreased the stock-out indicator by 3.8%.
Related articles and content
If you want to learn more about related topics such as e.g. pricing as well as simulation methodology you can have a look at the following SCDA articles:
- Link: Price optimization for maximum price and inventory control
- Link: Manufacturing simulation for plant design
- Link: Simulation-based capacity planning
- Link: Backlog simulation in FIFO production system
- Link: simmer R package applied to simulate simple receival inspection process
- Link: Visual Components financial KPI simulation
- Link: Simulation methods for SCM analysts
- Link: Monte-carlo simulation in R for warehouse location risk assessment
- Link: Parking lot simulator with simmer in R
Hi! My name is Oswaldo Almonacid and I am the editor for Latin America of this blog.
I am a supply chain consultant with particular interest in applying analytics-enabled solutions and operations research methods to help clients achieve superior supply chain performance. Currently, as a Project Leader at Black Andes Analytics, I have engaged in projects to optimize inventory management, demand forecasting and operations visibility in the health care, cosmetics, oil & gas, and CPG industries in the United States and Latin America.
I hold a master’s degree in Supply Chain Management from Massachusetts Institute of Technology (MIT). Moreover, I am currently pursuing a master’s degree in Analytics at Georgia Institute of Technology.
Leave a Reply