Simulation-Based Inventory
Planning for the Digital Supply
Chain Era White paper
1
Contents
01 INTRODUCTION .................................................................................................................... 01
02 WHY SIMULATION FOR INVENTORY PLANNING? ........................................................... 02
03 WHAT CHALLENGES CAN BE SOLVED ............................................................................... 04
Defining Policies ....................................................................................................................... 05
Safety Stock Estimation ............................................................................................................. 08
Multi-Echelon Inventory Management ........................................................................................ 10
Bullwhip Effect Quantification ................................................................................................... 13
04 CONCLUSION ....................................................................................................................... 17
05 ADDITIONAL RESOURCES .................................................................................................. 18
Introduction
Inventory management is arguably the most important, and For example, the bullwhip effect, where unpredictable demand
difficult, aspect of supply chain planning. For business today, swings proliferate across tiers, causing stock-outs and backorders.
hazards develop ever more quickly, and increasing global True multi-tier inventory optimization in the changing world
interdependency means the risks are greater. The impact on requires new technologies that allow analysts to take a holistic
inventory planning is significant. view of the end-to-end supply chain.
Today, there are more interdependencies in the world than ever. The 2017 MHI Annual Industry Report showed that more than half
Change is rapid, and it’s becoming harder to predict the outcomes of supply chain leaders consider predictive analytics, including
of all kinds of decision making. Inventory planning, being the key network modeling, a disruptive technology for the industry.
driver of supply chain flexibility and resilience, is probably the most Predictive network analytics is usually carried out using either
affected part of a supply chain analyst’s work. mathematical (analytical) modeling or dynamic simulation modeling.
The global digitalization of the supply chain is fundamentally Analytical calculations and optimization, being the most
transforming the industry, and leaders are having to adapt by widespread technologies in the supply chain world, have been
employing the most up-to-date technologies to improve network unable to provide solutions with the level of detail and precision
performance. For example, customer purchasing behavior has required to address the challenges identified above. That is why
changed significantly, and now clients are striving for faster dynamic simulation modeling has been gaining popularity among
response times at lower costs. The concept of omnichannel is leading professionals in recent years, and is increasingly becoming
radically changing the game in retail, making it harder to forecast mainstream, just as optimization did before.
demand at the point-of-sale and distribution locations, place the
right inventory, and define safety stocks, without freezing too much Inventory management is one of the most illustrative business
working capital. challenges that can be solved using simulation. Let’s have a look at
what advantages simulation provides, and how it can help analysts
Even single-echelon supply chains are now complex systems, and define the right inventory strategies.
multi-echelon networks are much harder to predict and manage.
1
More than digital plus traditional: A truly omnichannel customer experience. Bianchi et. al., McKinsey & Company, 2016.
01
WHY SIMULATION FOR INVENTORY PLANNING?
Why Simulation for Inventory Planning?
A lot of questions are left unresolved when planners only use
analytical methods. For example, when managing a multi-echelon
supply chain, a planner cannot adequately answer questions such as:
• What is the desired service level between the first and second
level DCs to achieve the target end-customer service level?
• How does lead time variability at upstream suppliers affect service
levels at the point of sale?
• How do you forecast demand for second-echelon DCs, and how
much should be ordered from the first echelon, considering lead
time uncertainty?
• How do you measure and manage the demand amplification
bullwhip effect that tends to appear in multi-tier supply chains?
• How do you achieve network operation visibility across all
echelons?
Inventory management requires taking into account a great deal
of operational complexity when conducting planning. Traditionally A static formula cannot forecast the real situation in a logistics
inventory policies have been defined based on analytical calculations. network for a given day because it does not capture the operational
dynamics. For example, there is no way to define exactly how much
Analytical formulas are used in logistics to cost-effectively manage of a product was ordered on a particular date. Static models just
facility inventory dynamics, satisfying varying demands in unpredictable can’t deal with such detail.
environments, like with stochastic lead times. However, this largely
spreadsheet-based approach is unnecessarily rigid and incomplete. Using formulas, a manager can try to guess what the safety stock or
Formulas help define rules, but you can only guess how a supply chain inventory policy should be. This means they cannot really see and
operates when using them. It’s hard to validate whether the suggested prove whether a formula is suitable for their network before a policy
policies would work efficiently under real-world conditions. is implemented in the real world.
02
WHY SIMULATION FOR INVENTORY PLANNING?
Using a trial-and-error approach may cause severe disruptions and Simulation is the only appropriate technology available today
huge losses before a proper solution is found. That's why leading for detailed inventory policy planning, for four main reasons:
companies need to plan operations in detail before
implementation. • First, it captures operations over time, allowing analysts to study
gradual step-by-step changes in a system, and to examine
Therefore, formulas need to be complemented with a tool that inventory levels for any given moment.
can verify their appropriacy – a dynamic simulation model, • Second, it has the ability to account for real-world randomness.
sometimes referred to as a digital twin. Variable factors such as delivery time and demand-volume can
affect supply chain resilience. Simulation can take this variability
Simulation allows analytical calculation results to come to life in into account for more precise forecasts.
a virtual model, and to be checked and verified, showing how to • Third, simulation provides analysts with the ability to calculate all
actually implement an analytical model recommendation in a real- the statistics they need. This makes it possible to evaluate the
world supply chain. This enables planners to continue supply chain performance of a supply chain from multiple angles when deciding
fine tuning for sustained profitability improvements. on an inventory policy.
• Lastly, simulation can capture the uniqueness of any supply chain,
representing the complexity and special traits of a company’s
business without compromise.
More about the benefits of dynamic simulation modeling can be
found in the supply chain technology overview white paper. Here,
we’ll take a deeper look at what problems simulation can solve in
inventory planning.
03
WHAT CHALLENGES CAN BE SOLVED
What Challenges
Can be Solved
There are four main issues that simulation can help analysts with in
inventory planning and management:
• Inventory policy testing – defining the right one and finding the
best-performing parameter sets for these policies.
• Safety stock estimation – taking into account operational and
disruptive risks.
• Multi-echelon inventory planning – considering stocks across the
whole supply chain, including multi-tier demand.
• Bullwhip effect quantification – caused by demand variability and
a lack of visibility across echelons.
Let’s take a closer look into each of them.
04
Defining Policies
05
DEFINING POLICIES
Simulation can help analysts define the right inventory policies. Whether By running these experiments, you can easily compare the results
it’s a Min-Max, RQ, or Material Requirements Planning (MRP) policy, you of multiple simulation runs, and thus identify what your policy’s
need to figure out what works best before implementation. Also, you parameters should be, like replenishment points, order amounts,
are likely to want to test your own, and more complex inventory policies. order aggregation periods, truck load policies (FTL/LTL), and more.
A good simulation-based supply chain tool can provide this ability. In-depth simulation tools also include the safety stock estimation
experiment which will be discussed later.
With simulation, an analyst can choose any policy and run a simulation to
see how a network operates under those rules. They can also apply their When evaluating policies and parameters, you may look at
existing policies for forecasting or stress-testing and see how resilient different types of statistics like service levels, total operating costs,
to disruption their current system is. When evaluating changes, like new cost-to-serve, cash-to-serve, EBITDA, and more. Because simulation
product introduction, demand spikes and drops, choosing a new can account for so many KPIs in the same model simultaneously,
supplier, or opening a new distribution center, the analyst can vividly analysts can thoroughly examine a model and make the right
see how this change will work on their distribution network. decision bearing in mind all the required statistics.
When developing a policy, analysts can conduct multiple experiments,
varying policy parameters to find a suitable combination. Common
types of simulation-based experiments are:
• Run comparison experiment – several simulations with different
input parameters are run, and the results compared. It allows you
to view how differences in the initial conditions, as well as random
parameter values, affect the outcome. You can use it when several
alternative inventory policy configurations need comparing.
• Variation experiment comprises a series of simulations based on
a single scenario, with one or several parameters being varied.
Variation experiments are used in two main cases:
• To see how changing a parameter value in a scenario affects the
supply chain.
• To see how a randomly changing value in a scenario, like a
delivery time, affects the supply chain.
06
DEFINING POLICIES
Planners in a U.S. industrial products manufacturing
company struggled with high inventory instability caused
by uncertainty in demand planning. The company needed
a new and improved inventory policy that was adaptive
enough to accommodate the unpredictability of a rapidly
changing environment. Using analytical calculations, the
planners came up with a complex inventory policy based on
the statistical process control (SPC) method. To evaluate this
policy’s performance in a risk-free environment, they built a
simulation model of their supply chain.
The model they built, using anyLogistix® software, included
an 11-product supply chain with production, distribution,
and supply echelons. It dynamically simulated the network’s
operations and allowed the planners to see how metrics were
influenced by different factors and ordering events over time.
First, they validated the model using their current MRP and
reorder point policies. After that, they introduced the new
Case study: SCP policy to see how it affected stock and service levels.
They were able to fine tune policy parameters to find the
industrial products most profitable combinations. Experiments showed the
manufacturer resulting policy would reduce volatility and stock, allowing
the company to implement changes in real life without taking
risks of stock-outs and unfulfilled demand.
07
Safety Stock Estimation
08
SAFETY STOCK ESTIMATION
Right-sizing safety stock is one of the most
challenging tasks in inventory management.
It requires accounting for the dynamics of
network operations, as well as operational
and disruptive risks, demand variability across
multiple echelons, and the unpredictability of
omnichannel.
Simple safety stock formulas just won’t work
for real-life complex, lean, and agile networks.
Averaging complexity is just putting operations
at risk. On the contrary, simulations can
represent the supply chain operations as they
are, thus providing the technical ability to
— Inventory behavior in the ideal world and in the reality.
define the right safety stock for the given
conditions. With a simulation, it is possible to allow your stock to fall below zero, to see how much
In an ideal world, inventory is always kept low, product you would actually need, and get a time progression chart like the one above.
and you never run out of stock. Imagine a If your target service level is 100%, then set your safety stock to the lowest fall below
simple Min-Max (or S, s) policy, where upon zero – in this case, 100 units. If it’s 95%, cut off the lowest 5% of cases, and so on.
reaching reordering point s, you place an order A time progression chart like this can only be obtained with the help of dynamic
to return to quantity S. Simple inventory simulation. Only simulation can handle the actual course of time. Analysts can achieve
forecasts and plans are often like the green line inventory dynamics visibility by tracking the stock level live and delivering a lean,
on the chart on the right – you hit zero resilient, and optimal supply chain.
inventory at the moment when the
replenishment comes in. But in fact, the A robust supply chain analysis tool needs to have a built-in capability to carry out
situation usually plays out in a different way such safety stock estimation experiments for each facility and product group.
(the black line), and, following these rules, you Analysts should also be able to calculate service levels by orders, products, revenue,
often experience stock-outs. and various types of ELT service levels.
09
Multi-Echelon Inventory Management
10
MULTI-ECHELON INVENTORY MANAGEMENT
When conducting multi-echelon inventory management, the goal A simulation model can dynamically represent the operations of all
is to deliver the desired end-customer service levels with minimum echelons exchanging products, cash, and information as a complete
inventory investment across the various echelons. system, with a single source of demand from end-customers,
avoiding multiple independent demand forecasts in each echelon.
Dynamic simulation modeling is the way to forecast true inventory It can account for:
costs. Multi-echelon means multiple interdependencies in the
network that may not be obvious at first sight. At times there are • All randomly varying lead times in and between the echelons.
many independent parties involved in a value chain, which makes it • Differentiated service levels for DCs on all echelons and for all
even more difficult to manage. To be efficient, you must provide end-customers.
streamlined communication, clear visibility, and coordination • Different replenishment strategies at all sites and tiers.
among tiers. A single source of information and planning is
In addition, with top-to-bottom visibility, the logistics network
essential to meet this requirement.
allows analysts to monitor and manage the bullwhip effect.
An end-to-end supply chain simulation model can act as a digital
twin, providing the required visibility into multiple echelon
operations. To work well, such a simulation model should employ
precise operational data from all of the parties involved. The
effort pays off in precise forecasting, lower operational costs, and
greater risk resilience for all echelons.
11
MULTI-ECHELON INVENTORY MANAGEMENT
The company was optimizing the manufacturing supply chain
for two products. Each of them had their final manufacturing
stages in the same factory but had differing logistics networks
in their preceding stages. The bills of materials within the
scope of the project consisted of 14 products, across six
echelons (consecutive production sites), at different locations
in Asia, Europe, and North America.
The manufacturer needed to control inventory and service
levels, as well as maximize factory production capacity
utilization. The simulation model developed in anyLogistix®
software allowed the planners to gain visibility into the supply
chain and discover the effects of policy changes. The company
was able to improve operational performance by reducing
stock for certain products. The solution was developed
through experimenting with different inventory policies and
their parameters.
Case study:
pharmaceutical
manufacturer
12
Bullwhip Effect Quantification
13
BULLWHIP EFFECT QUANTIFICATION
The bullwhip effect in supply chains, caused by demand Customer demand volatility has become one of the biggest risk
variability, inventory planning rules, and sourcing policies, is a drivers in the supply chain world. The bullwhip effect makes this
phenomenon that is challenging to measure. The bullwhip effect problem even more threatening. Some of the operational reasons
appears when demand swings are heavily amplified as they move for the effect are: poor forecasts, huge lead time variability, human
upstream through the network. For example, the demand at the factors (panic ordering), and the lack of visibility across the demand
manufacturing stage is sometimes several times higher than chain, as well as a lack of coordination and communication between
the demand at the point of sale. This can lead to stock-outs and tiers. To minimize its negative consequences, all parties across the
backorders across the entire network. value chain must understand how the system operates.
A simulation model, acting as a digital twin of the supply chain,
can help with this. All the parties involved, from all tiers, can align
production and inventory plans with point-of-sale demand.
— Bullwhip effect in the classic Beer Distribution Game supply chain.
14
BULLWHIP EFFECT QUANTIFICATION
A digital twin can reproduce the following:
• Demand variability, including random peaks occurring with
probabilities and seasonality.
• Multi-echelon inventory policies with actual values, operating
over time.
• Transportation policies and lead times, taking into account
uncertainties and in-transit inventory.
• Manufacturing and supplier lead time variability, taking into account
the inside four walls production operations and supplier reliability.
This visibility not only helps create transparency across tiers, but
also helps identify and handle the causes of the bullwhip effect,
such as inefficient lead times.
With simulation, the bullwhip effect can be quantified. This is done
by calculating a demand variability amplification coefficient – a single
metric providing a holistic view of a whole network’s operations.
A suitable simulation tool must be capable of reproducing supply
chain operations for further analysis, as well as having a toolset to
observe and, more importantly, quantify the bullwhip effect.
15
BULLWHIP EFFECT QUANTIFICATION
Infineon, a large semiconductor manufacturer in Germany,
wanted to investigate the bullwhip effect in their market to
help decrease expenses and better forecast market behavior.
They built a simulation model of their supply chain, from raw
materials to consumer market. Specifically, they wanted to
discover the connection between industry demand fluctuation
and the demand they received from direct customers, across
six tiers of the network.
The model realistically reflected the anxiety that appeared
in the chain, including under- and over-ordering behavior. It
helped the company analyze fluctuations emerging in the
market, estimate and quantify the consequences of the
bullwhip effect and the amplification of demand along the
supply chain1.
1
Based on materials from https://www.anylogic.com/bullwhip-effect-in-semiconductor-supply-chain
Case study:
semiconductor
equipment producer
16
CONCLUSION
Conclusion
In today’s supply chains, analysts have to consider all network specialized supply chain software tools on the market have the
complexities when making strategic managerial decisions on capability for dynamic simulation, and even fewer have made it really
inventory policies. The technology employed should be able to easy to use, without the need for programming skills from the users.
reflect all of the unique traits in a company’s network, including
Advanced supply chain simulation tools have prebuilt functionality
whether the planners are targeting for a lean or agile supply chain.
to streamline experimentation on inventory policies for easily
Dynamic simulation modeling is becoming a go-to technology when testing multiple simulation scenarios, comparing them, and
needing to analyze a multi-tier supply chain in operation, verify choosing the best parameter sets. Also, a good tool includes
new inventory policies before implementation, define safety stock problem-oriented, prebuilt simulation experiments for inventory
volumes, or estimate the bullwhip effect. No other technology planning. For example, anyLogistix® supply chain software includes
used for logistics planning, such as analytical methods or network specialized simulation-based experiments for safety stock
optimization, is capable of dealing with these problems. Simulation estimation and bullwhip effect quantification.
is the technology of choice because it is the only way to create
Dynamic simulation modeling constitutes a new generation of
a supply chain model that can take into account dynamics and
predictive modeling and analytics. Simulation uniquely allows
randomness, along with all of the necessary details and metrics.
the creation of a supply chain digital twin, taking into account
Simulation is already a mature technology. Vendors report actual network behavior over time, and calculating the statistics
that most Fortune 500 companies are using general-purpose live, as time progresses. More and more companies are adopting
simulation tools to solve a variety of problems, including logistics, this technology, and industry leaders and challengers should have
manufacturing, finance, and marketing, and the number simulation technology in their inventory analytics toolset.
of adopters is ever growing. Still, despite this, only a few
17
Additional resources
• anyLogistix Webinar
• anyLogistix Demo Video – tutorials
• anyLogistix PLE – free version for self-study and education
• Tutorial book
• White papers
• Case studies
18
Contacts
The AnyLogic Company
Choose your local office >>
[email protected]
19
