Logistics Network
Design
• Introduction
• Cost Relationships
• Methodology
Identifying the appropriate strategy that are best suited for a
variety of customer and product characteristics is crucial to the
success or failure of supply chain mission
Main reason to develop a Logistics Network
The need to provide an effective service to the customer,
while minimizing the cost of that service
A Logistics Network Design Optimization has to provide answers
to following questions:
• How many distribution centres (DCs) do we need to meet the customer expectations
(e.g. lead times), and serve the Road to Market strategy?
• Where should the DCs be located?
• What are the required storage capacities of the DCs, for the coming years?
• Which customers should be serviced by each DC?
• Should we close / move / outsource existing DCs?
• What is the optimal TO BE scenario? What are the quantitative & qualitative benefits
vs. the AS IS situation?
• What would be the optimal high-level layout & automation level of own DCs?
• What are the specifications regarding t° control, safety…?
• What is the agreed implementation plan?
• What are the quick wins?
Designing for a sustainable future:
how to model carbon costs in network design?
• Traditionally, network design focused on optimizing costs and service levels
• More recently, designing for resilience and agility has become more prominent
• Going forward, another dimension will require more attention: designing for
sustainability and responsible business
• Modeling carbon costs can be a helpful starting point to understand how climate change targets
will impact your supply chain. If your model data is well-designed, your life will be a lot easier
• More functionality will progressively be released to support this in the future
Different types of Distribution Centres (DCs)
• finished goods DCs/warehouses – these hold the stock from factories
• Distribution Centres, which might be central (CDC), regional (RDC), national (NDC) or
local (LDC) DCs – all of these will hold stock to a greater or lesser extent
• trans-shipment sites or stockless, transit or cross-docking DCs. These DCs do not
hold stock, but act as intermediate points in the distribution operation for the
transfer of goods and picked orders to customers
• seasonal stockholding sites
• overflow sites
Logistics Network Design & Sourcing analysis
• The main elements of a logistics network planning study revolve around the location
and use of the distribution centres (DCs)
• Before trying to determine the most appropriate number & location of DCs, it is also
necessary to ensure that there is an efficient flow of products from source to final
destination. This assessment of the different patterns of product flows is known as
sourcing analysis.
Logistics Network
Design
• Introduction
• Cost Relationships
• Methodology
Cost Relationships
Logistics network and DC location strategies are aimed at
establishing the most appropriate blend of storage and transport,
at a given customer service level
Storage Costs
The interrelationship of the
different distribution Inventory Holding Costs
elements and their associated
costs thus provides the basis Transport costs (primary & secondary)
for decision making Information Systems Costs
Cost Relationships: Storage Costs
Storage costs considered here:
• Building
• Building services
• Labour
• Equipment
• Management/supervision
As the number of DCs in a
distribution network increases,
Relationship between number of depots then the total storage (DC) cost
and total storage cost will also increase
Cost Relationships: Inventory Holding Costs
Inventory Holding Costs considered here:
• Capital cost
• Service cost (stock mgmt. + insurance)
• Risk costs (damaged, obsolete..)
• Storage costs (see previous slide)
As the number of DCs in a
distribution network increases, then the
total Inventory Holding Costs will also
Inventory holding costs in relation to the
number of depots
increase
Cost Relationships: Primary Transport Costs
The primary transport element is the
supply of products (ie in full loads) to
the DCs from the central finished
goods warehouse or production point
Often, the effect is not a particularly
large one, but it does result in an
increase in primary transport costs as
Primary transport costs in relation the number of DCs increases
to the number of depots
Cost Relationships: Secondary/Delivery Transport Costs (1/2)
Delivery transport = the delivering of orders from the DC to the customer.
This cost is essentially dependent on the distance that has to be travelled.
Delivery transport distance has 2 components: stem & drop distances
‘Drop’ distance = the distance travelled once a drop or delivery zone has
been reached (starting from the first ‘drop’ point and finishing at the final
‘drop’ point)
Drop distance will not change if the number of DCs change because the
First Drop Last Drop delivery zone(s) will remain the same
‘Stem’ distance = the distance to and from a delivery zone (from the DC
outwards to the first ‘drop’ point and return from the final ‘drop’ point back
to the DC)
Stem distance to any particular delivery zone will depend on the location of
the DC. If the DC is changed then the stem distance will also change (it will
increase if there are fewer DCs and will decrease if there are more DCs).
DC
Cost Relationships: Secondary/Delivery Transport Costs (2/2)
The greater the number of DCs,
the less the stem distance
➔ Lower Delivery costs
Delivery transport costs in relation
to the number of depots
Long stem Short stems
Cost Relationships: Combined Transport Costs (Primary + Secondary)
The overall effect of combining the
two transport costs is that total
transport costs will reduce,
the greater the number of sites that
there are in the system
Combined/total transport costs in
relation to the number of depots
Cost Relationships: Information System Costs
System costs are less easy to represent graphically
because:
• of the fast rate of change of information systems
• costs can vary considerably dependent on the
level of technology introduced
The higher the number of depots in a logistics
operation, the higher information system costs are
Information system costs in relation to likely to be
the number of depots
Cost Relationships: Total Logistics Cost
The minimum point on the overall logistics cost
curve represents this lowest-cost solution
The results, in practice, will depend on a
number of factors:
• product type
• geographic area of demand, demographics
and population density
• service level required
The relationship between total and • availability of suitable buildings
functional logistics costs as the number • Criminality level in some regions
of depots in a network changes
Trade-off analysis: example of #DCs reduction
Trade-off analysis showing that a change in configuration can lead to a reduction
in total logistics cost while some cost elements increase and others reduce
Logistics Network
Design
• Introduction
• Cost Relationships
• Methodology
Methodology: introduction
• Determining the optimal number, size and location of DCs is a complex process
• The vast majority of location studies are undertaken when a company already
has an established logistics network
• The majority of existing systems have just evolved on the way that the company
has evolved. This may be a steady growth (or decline), or may be in short steps
or large leaps as mergers and takeovers occur
• Probably, the most common reason why logistics networks are out of balance is
that of inertia, because of the great amount of work and effort required to
make changes
• It needs a forward-looking management or a particularly significant change for a
company to undertake a large-scale study of this nature.
2 Internal Factors
• Importance will vary from
industry to industry
• Qualitative factors (product
range extensions, service
offered to customers…)
• Quantitative factors (growth,
acquisitions…)
1 External Factors
• transport mode availability
• infrastructure changes (eg
new roads, rail links, etc)
• regulatory changes (transport
legislation, customs
regulations, etc)
• environmental impacts
• industry trends
• Geopolitical situation
• …
External
Factor
Internal
Factor
3 AS IS situation
• Mathematical description of the main
current product flows and costs of the
logistics operation
• The resultant mathematical model will
be used as the basis against which all
alternative solutions are measured
later in the study, so data accuracy is
crucial
It is a very detailed and necessarily time-
consuming process but is essential to the
study
AS IS
Always make visual
representations
Logistics network flow diagram, showing
some examples of major flows and costs
4 As Is DATA Collection
• data that will be used as the basis for
determining the current situation and
for the subsequent logistics modelling
and analysis that will underpin the
entire strategy planning process
Be well aligned on the:
• Unit of measure
• Product groups
• Customer classification
• Time periods
AS IS • AS IS Data collection is often the largest part of a study
• There are always problems in finding and obtaining the
data data and information required
Typical examples of descriptive data include: Examples of quantitative data are:
• product groups • major product flows
• own and bought-in sourcing locations • transport modal split for the major flows
• number and type of sites and facilities • demand by region, by major product group, by
• major transport modes customer type, etc
• handling systems used • market segmentation
• unit load types (pallet, truck, sea containers…) • customer service goals and achievements
• own versus third-party operations • carrier analysis
• main customer groups • inventory holding profile
• customer service levels • product profile
• logistics information systems • customer profile
• expected growth rates
• planned future expansion requirements
5 Customer service analysis
• customer service requirements
provide a key input to any logistics
network planning study
• Appropriate customer service levels
that need to be achieved must be
identified.
• These are essential ingredients to
help determine the type of logistics
structure that should be developed
• There is a vast difference between
offering a low-cost (high
productivity) solution compared to a
high-cost (value-added) solution, and
the ideal logistics structure will vary
accordingly
• A logistics network is likely to serve
several different markets and needs
to be designed accordingly in order
to satisfy all of the different
customer service requirements
6 Determine logistics objectives
Execution of steps 0 to 5 should
allow for clear logistics objectives
to be determined
7 Logistics options
• generate a long list of alternative
scenarios that would be worth
considering for modelling and
assessment
• then by fairly simple assessment
determine which of them, or which
combination, may be feasible in the
context of the planning horizon (=
short list of potential TO BE scenarios)
TO BE
scenarios
• First diverge before converging
• 1st part = Divergent Thinking
• Organize a creative session to brainstorm on all possible
TO BE solutions
• No bad ideas, no constraints, think out of the box
• 2nd part = Convergent Thinking
• Eliminate less relevant TO BE scenarios, based on high
level analysis & common sense
• Confirm limited set of TO BE scenarios (+/-3), to be Deliverables:
assessed in details • Realistic set of most relevant
TO BE scenarios
Some ideas for TO BE scenarios (1/2):
TO BE
• Relax transport constraints
• This will show you opportunities for allocating customers to different service scenarios
locations to optimize costs. These are often “low hanging fruit” as they require
little or no investment to execute.
• Relax DC capacity constraints
• Relaxing DC capacity constraints may highlight opportunities to shift volumes between locations
to save costs. It will also reveal the quantum of additional capacity required at DCs to make these
shifts, and whether it is worth the CAPEX investment to do this compared to the OPEX savings
that can be achieved.
• Relax supplier and production constraints
• This will show you shifts in supplier and production volumes that may deliver a cost improvement.
In some cases, supplier and production shifts can be done easily without massive investment. In
other cases, the opposite is true. But either way, this scenario will highlight the quantum of the
shift and the expected OPEX benefits, which can then be weighed up against the CAPEX required.
• Allow the model to close locations
• This is a high-impact decision. This scenario helps you identify the locations that could be
considered for closing in order to save costs
Some ideas for TO BE scenarios (2/2):
TO BE
• Explore CoG (Centre of Gravity) candidate locations scenarios
• A Center of Gravity analysis helps you compare the current
network to a hypothetical “blue sky” network. It helps you to
identify some candidate locations.
• Add Lead time constraints
• The optimization will give you the “best cost” network but still
respect lead time (drive time) constraints. This will result in a
balance between cost and service levels that may be a better
solution than just optimizing costs.
• Run some demand flex scenarios
• Running optimizations against the base case network can reveal
many insights. You should also optimize for the future by
running some scenarios against projected future demand. These
projections may not be accurate, but you could run high growth
/ low growth / expected growth scenarios to put some
confidence bounds around your forecast.
8 Determine modelling approach
• The combinatorial problem of considering
all products, made at all sites, shipped via
all modes, to all customers, via all DCs is
simply prohibitive
• a rough-cut approach can be adopted for
strategic logistics planning, and can be
described as trying to establish the
‘economic footprint’, roughly the physical
boundary that the facility can serve.
• The economies of scale of production, the
customer service requirement and the
logistics cost are all considered to give an
optimum factory size and radius of
operation, hence the economic footprint
• A means of ‘rough-cut modelling’, for the whole of
a logistics operation, is to use sourcing models
• Costs of raw materials, production rates and
capacities, together with approximate logistic costs
across a geographical area, are used to calculate
the trade-off between the major elements
Economic
footprint
Small Medium Large
9 Sourcing models
• With multiple products from multiple
sources, it is only too easy to assume that
the lowest-cost solution is to source each
market from the closest available plant
with available capacity
• In some situations this is true, but if plants
have significant changeover (set-up) times
it may be more cost-effective to have long
production runs, high inventory and high
transport costs
➔ the first step in rationalizing a logistics
system is to investigate optimal sourcing
patterns
The optimum sourcing solution will define
Sourcing the type of the required DC structure
Models
Order flow Stores
Physical Flow Customers
DCs required No DC required
Sourcing
Models
• Linear programming is a mathematical technique that finds the minimum
cost or maximum revenue solution to a product sourcing problem
• All available sources are described with capacities, changeover penalties
and raw material costs.
• Approximate logistics costs from sources to markets are defined as linear
cost functions
• Under any given demand scenario the technique is able to identify the
optimum solution for the sourcing of products
[Link]
10 Distribution Centre location modelling
• The output from a sourcing study (step 9) is
the optimized major product flows from
source points to final customer
• The next stage is to take these flows and to
develop the most cost effective logistics
solution in terms of the most appropriate
number, type and location of DCs,
transport mode, etc.
➔ The overall trade-offs of the supply chain
are considered and assessed during the
sourcing study, and a preliminary sourcing
allocation is made.
The detailed logistics of modes, rates and site
structure can then be considered using a DC
location study.
Cost trade-off analysis can be used with
several different approaches
DC location The most common approach to determining DC location solutions
is to use a logistics strategy simulation or optimization model.
modelling A variety of software programs is available on the market.
Other potential approaches:
• Mathematical programming, such as linear programming. It is not always the most adequate
approach for this step
• Heuristic methods are determined by using experience and common sense to reject unlikely solutions
from the outset. This type of approach is often valid for DC location problems, because there are likely
to be a number of locations that are totally inappropriate
• Simulation is a widely used operational research technique. It is not an optimizing technique, so does
not produce the ‘best’ answer. Simulation models allow for various ‘what if’ questions to be asked to
test alternative strategies
• Certain practical approaches can also be used – particularly to help simplify the analytical approach:
represent the network on a map / Clustering techniques / use a spreadsheet-based approach
11 Determine base case
= step 1 of DC location study
= build & validate the model that will be used
to simulate the different options, to make sure
that the costs are predicted with reasonable
accuracy
Base Case Modeling the base case to build a digital supply chain twin
Although it can be work-intensive to prepare a base case, it is necessary for two key reasons:
• Validating that the model represents reality
• Providing a reference point to measure optimization opportunities
The base case provides the foundation for a strategic, digital supply chain twin.
Once this digital representation of the supply chain is in place, it becomes easy to run multiple what-if
scenarios. We need to ensure confidence (both for yourself and for the business owners) that the model
actually represents reality
Without a base case, there is no way of providing the estimated benefits.
Base Case Tips and tricks for network design data preparation
1. Keep it simple: Remember that this is strategic modeling
2. Remove the noise (samples, returns…)
3. Aggregate the products: modeling at the SKU level rarely makes sense for network design
4. Aggregate the customers: Consider aggregating demand nodes at the city or zip code level
5. Choose the Units of Measure with care (kg, pallets…). Keeping it simple is the goal.
6. Use capacity constraint assumptions at production facilities, suppliers and DCs
7. Transport rate structures can be very detailed and complex ➔ simplify them (e.g. calculate average
€/ton cost per geography to approximate lane rates for lanes that are not currently in use)
8. Avoid multi-periods models, unless we need the complexity
• If you are worried about seasonality, consider running a peak month and an average month as 2 separate
single period datasets, instead of running a multi-period model with 12 months
9. Model Bill of Materials (BOMs) only when necessary
[Link] the business rules (storage & transport conditions, Regulatory requirements, Political and trade
rules)
Base Case Data collection tips: where can you find it? (1/4)
1. Base case data
• Is typically obtained from the ERP system or BI data warehouse
• Select a representative period (typically one year of history) without any major deviations from
the norm
• The data should allow to aggregate flows summarized by location, to location, mode of transport
and product group – depending on what will be used in the model.
• Try to obtain flows for primary, secondary and inter-resource transport.
• Try to obtain a dataset at a lower level of detail that can be aggregated upwards, rather than pre-
aggregated data. This will allow to re-visit the data aggregation assumptions if necessary. We can
always aggregate upwards from a detailed dataset, but aggregating downwards will usually
require a different dataset.
Base Case Data collection tips: where can you find it? (2/4)
2. Geographic data
• Your model will need geocodes (latitude and longitude) for each node, including demand points,
DC locations, supplier locations, etc
• This data may be held as master data in your ERP or routing and scheduling system. If not, you
should obtain address data (country/state/ city/zip code/street address)
3. Capacity constraints
Capacity constraint data is not readily available in transactional or ERP systems. Typically, this
information will have to be supplied by:
• production department (production constraints)
• warehouse manager (DC constraints)
• logistics department (logistics constraints)
Base Case Data collection tips: where can you find it? (3/4)
4. Demand Forecasts
• The base case data reflects what happened last year, but in network design we should take a
forward view to design for the future and not the past. ➔ we should run some scenarios with
demand adjusted for the anticipated future.
• The horizon will depend on what we are trying to achieve:
o a few months to model short-term risk scenarios
o 2-5 years for DC capacity questions
o 10 to 20 years for heavy industry with major CAPEX investments.
• This data is typically sourced from the business, and is not in the ERP.
Base Case Data collection tips: where can you find it? (3/4)
5. Costs
• The model will require cost estimates at the appropriate level for network design.
• This data is typically sourced from the finance department
o Supply and production costs
o DC fixed and variable costs
o Transport costs (primary, inter-resource and secondary)
• Logistics managers could also be another source of information. Often, they have done the
translation already, and have cost numbers at hand that can be used for validation against the
numbers coming from finance.
• Transport costs can be particularly difficult to define, as they can be very detailed and complex.
We’ll need to aggregate and simplify them for network design purposes.
• Transport costs can be obtained in a creative way from a mix of the following sources:
• The finance department / Logistics managers / routing and scheduling tool / TMS / ERP / our
3PL provider / Collaborative external data sources (Xeneta, Transporeon…)
12 Analyse, cost and compare options
= step 2 of DC location study
= test of a variety of alternative options
13 Evaluate results against logistics and
business strategy
Assess the impact of the best model(s), that
have been identified, on the total business
strategy ➔ impact on
• Capital costs
• Operating costs
• Customer service
Use a weighted mix of quantitative (costs) and
qualitative (service, flexibility, risks,
implementation effort) criteria’s to assess the
different strategic options identified
Evaluate
results
Example of part of a qualitative
assessment used for a study
14 Implementation
When an appropriate option has been finally
agreed and the remaining distribution
requirements have been determined,
then the process of implementation can start.
This should follow the standard business
practices used for any major project
implementation plan.
In addition, suitable cost and performance
monitoring processes should be set up and
used
Who could perform a Logistics Network design Study?
Criteria’s:
• Resources (people)
• Time
• Skills
• Internal vs. Consultancy costs
• Simulation & Optimization
software need/availability
• Independence
Representative Vendors in Supply Chain Network Design Tools
Main source Other sources:
• AIMMS-Implementing-Supply-Chain-Network-Design-Guide_2021
• [Link]
• [Link]
hamza-cpim-scm
• [Link]
• Chapter 9:
logistics costs &
Trade-off analysis
• Chapter 10:
logistics network
planning
