Distribution Network Design PDF

Preview

Summary

DISTRIBUTION NETWORK DESIGN **studia insieme alla parte sulla dispensa The objective of the second part of the course is combine the elements studied in the first part. In order to analyze the distribution network design, we will analyze: A. The principles (no quantitative models) 1. Problem setting...

Description

DISTRIBUTION NETWORK DESIGN **studia insieme alla parte sulla dispensa The objective of the second part of the course is combine the elements studied in the first part. In order to analyze the distribution network design, we will analyze: A. The principles (no quantitative models) 1. Problem setting & decision framework 2. ‘’Distribution problem’’, how to describe a distribution problem 3. Typologies of distribution networks 4. Design principles 5. Examples B. Methodologies & Models

A1 – Problem setting Logistics network is a combination of nodes and arc, connections. There are two main kind of network • Supply/procurement network • Distribution network The main differences between these two kinds of network are:

Supply procurement network (inbound supply chain) Distribution network (outbound supply chain)

Product Flow of raw material and components

Structure Convergent*

Flow of finished products

Divergent*

*Convergent structure, from many points of supply to few points of the supply chain (take resources to the plant) *Divergent structure, from few points of the supply to many points (from the plant to the different shops) In our course we will focus on distribution network, it’s important note that the 100% of the principles can be easily adapted to the supply network.

The problem Point of origin

Structure of the network

Point of destination

The main issue is understanding how to design the network to connect the points of origin with the destination ones. In order to do that we have to consider these influencing factors: •

Scope of the system you are considering, for example, we could have: Network of manufacturer that serve distributors Network of manufacturer that serve final costumer Network of retailer Network of manufacturer that serve retailers If you don’t define the scope you cannot understand which the points of origin and the points of destination are à you cannot design the network because you don’t understand the perspective.

•

Sourcing strategy + marketing/sales strategy Sourcing strategy: You produce directly in your plant or you use third plant? This change the point of origin. Marketing/sales strategy: how you serve the market? Indirectly with distributors or you serve directly the market, directly the final customer? This change the points of destination. Without this point you don’t understand the elements of the problem

•

Type of products What are the most important characteristics that influence the structure of the network? o Obsolescence/perishability. o The value of the products (value density, how many euro per kilos), affect the objective function because if we have high value density products, I will expect that the inventory cost are more than transportation and vice versa if we have low value density products. o Variety (product range), is a driver of complexity. . . . The objective function What you want to optimize? Logistics costs + service level

•

o You can minimize logistics costs subject to service level constrains (for example no more than 3 days, and you can do that for different service levels) Doing that you understand the logistics – service level curve

o Maximize service level subject to logistics costs constrains o Minimize (logistics costs + cost of ‘’poor service’’) Poor service cost is the cost for no give the service for example the stock out cost

Structure of the network The main dimension that define the structure of the network are: • Number and typology of nodes • The transport network, how the nodes are connected • The transport modes, road-rail-plane-ship • Inventory policy, where and how you keep your inventories This is the problem setting, now we will spoke about the decision framework

Decision framework The decision framework includes three decision: • Description of the problem – description problem profile (A2) • Understand the possible solutions – typologies of distribution networks (A3) • Understand the connection between these two points – Design principles (A4)

A2 – Understanding how to describe the distribution problem profile Understand the point of origin in term of • The number • The locations (area in which the flow start, are concentrate in one state or divided in many states?) • Specialization (are focused in which product? If we have some factories and the product range include 10 product families. How are these points of origins specialized? The factory one could be specialized in family 3 and 6 and so on. The portion of the product range that are focused. The consequence for the network design is that if an order includes all the family there must be a consolidation point in which we collect all the products. Understand the point of destination in term of • The number • The location • Order size, what the size of the typical order from the costumer. For example, fingle item or cartons? Understand the products in term of • The variety (product range) • The density of value • The obsolescence risks

Understand the service (target values of SL1, SL2, SL3...) in term of • Order cycle time • Punctuality (time window that the customers require) . . . Knowing that we have a good description of the problem.

A3 – Typologies of distribution networks Distribution networks can be classified on the basis of: Number of levels or echelons of nodes (depth of the network) • Direct delivery. No levels of nodes between point of origins and destination • One echelon network. There is one level of echelon. • Two echelon networks. There are two level. • Three echelon networks It’s clear that adding level of nodes in the network means increase the cost and complexity of the structure. This means that we have to optimize the number of levels. Number of nodes per echelon (width of the network) • Narrow distribution network (1, 2 … a few) • Capillar, wide distribution network (many) The performance of the network depends a lot on these two dimensions. Typology of nodes • Warehouses /depots/fulfillment centers, nodes in which you have inventories (you can pick orders there) • Transit points, nodes without inventories (you cannot pick order there)

A4 – Design principles Now we have to match A2 with A3. For which distribution problem profile, you suggest for example one echelon network? This is what we want to do in this point. Now we try to compare different typologies of distribution:

1-echelons network vs direct delivery What are the key possible functions of the first level in the network? What is the role of the first level in the network? The base case is direct delivery. In the direct delivery the inventories must be in the point of origin.

If you have one level of nodes the possible benefits, potential functions are: • Order consolidation, consolidate the order when the point of origin are specialized. More are specialized more the level of nodes is important. High specialization means that each point of origin can give just a portion of the product range. In this way we increase the completeness of the order. à You are exploiting this point when there is high specialization in the points of origin. • Reduce the order cycle time (with the assumption that the first level is closer to the customers than the point of origin and that the first level consists in warehouses), be able to fulfill the order in lower time because we are closer to the customer. àThis benefit is more important when the points of origin are far (geographically and in term of time) from the customers. Order cycle time target values, if it should be low, we must be very close to the market. • Reduce transportation costs, it’s important underline that using one level the travel increases because we don’t have a linear travel. We are triangulating the travel. You are reducing the number of connections indeed you are changing the number of arcs in the network. If we have direct delivery, M point of origin and N point of destination, high specialization and the customer order everything. We have M x N connections. Instead if we have one echelons network with one warehouse in which you consolidate all the orders. We have M+N connections. MxN

M+N

M=10 N=1000 Direct delivery, we need 10000 1-echelon network, we need 1010 We are reducing by one order of magnitude the number of connections. Not only we are reducing the number of connections, but we are also increasing the saturation of the trucks because we are concentrating the flow in a fewer connection, and we have transport consolidation. These three factors decrease the cost of transportation. This benefit increases when à we increase the number of points of origin and destination, the complexity of the network. à the transportation costs are an important part of the overall logistic cost. So, we can star from the description of the problem and understand if we should use 1-echelon network or direct delivery.

Examples 1 – Bassetti, 3M, Unilever – 1echelon network Distribution problem profile Point of origin

-Number -Location -Specialization

Products

-Product range -Density of value -Risk of obsolescence

Point of destination

-Number -Location -Order size

Service

-Order cycle time

The point of origin are manuf. plants

Bassetti -10 -ITA + East Eu -Are this point of origin specialized in one product of the product range or they can produce all the product rangeà focus factories

3M Manufacturing -50 plants -EU -focused

-10000 - 1 euro per kilo low 10 euro per kilo medium 100 euro per kilo high à high -medium high because a big portion of the product range change each season in term of color size etc. The points of destination are mainly shops.

Type of network

-5000 -EU -some item, some cartoons, some cartoon, full truck à pallets/cartoons

Manufacturing plants

-50000 -high -medium

Intermediate customers, other manufactures or wholesalers

-some weeks 1-echelon distribution network with one warehouse

-10000 -EU -pallets

Unilever -20 -EU -general type of plants, low level of technology is needed to produce this type of product

-300 -medium, low 1 or 10 euro per kiloà the incidence of transportation cost is high, for this reason they produce only in Europe e not in the world. They should be near to the customers -low Mass market retailer (distribution center of the supermarkets)

-one-two weeks 1-echelon distribution network with 5 warehouses

-5000,10000 -EU -full truck load or fractions of full truck load -some days 1-echelon network with more than 10 distribution networks

Why they decide to use 1-echelon network? Bassetti Potential functions

3M

Unilever

Order completeness

Important because of the high specialization

Important because of the high specialization

Not important

Reduce order cycle time

Not important

Important (it depends on the type of customers)

Important

Transportations cost

important

Important, less than uniliver. Because we have a lot of point of origin and destinations à complexity

Important because of the characteristics of the product (medium low density of value that means high importance of the transportation cost on the overall logistic cost) and the complexity of the network

Looking at the potential function we find almost one factor that is important for the company.

Principles 1.1 Level 1 of the network has three main potential functions: • Increasing order completeness (order consolidation) • Reducing order cycle time • Reducing transportation cost If one of this function is very important, level 1 should be consider as a real option. Principles 0 (corollary of principles 1.1) Direct delivery having not echelons is an interesting solution when no one of the level one potential function is important. Why did they use a different number of nodes per echelon? Leadership cost point of view: Clearly when you increase the number of nodes at the first level the service level increases but also the logistics cost. In particular we have an increase of the warehousing costs and the inventory costs (because we have to split the inventories in different warehouse and consequently, we have to create a safety stock for each new warehouse à the overall safety stock increase). The transportation cost tends to reduce till to a certain point because more warehouse means more opportunities to manage the transportation flows. With one warehouse we have to concentrate all the flow in one warehouse, it’s easier and convenient in term of transportation cost put a warehouse in each specific country. The logistic cost has a typical U-shape, we can find the optimum. The optimum can change if the component (transport and warehouse have different importance). When warehousing and inventory cost are more important than transportation one, we will use few warehouses. Instead, we move to more warehouse if the transportations costs are more important than the warehouse ones.

Uniliver, for them transportation cost is very important. Instead the warehousing cost in not so important (low value and low obsolescence) à they use more warehouses Bassetti, big product range, high value density, high obsolescence transportation cost are not so important, are important instead the warehouse cost à in the opposite side of Uniliver 3M, the problem is very similar to Bassetti one but now we are talking only about cost. We have to speak about the service level constrains. Service level point of view: Bassetti, no service level constrains àconsistent with one warehouse Uniliver, high service level constrains à consistent with many warehouse 3M, we are between the optimization of cost (network like Bassetti) and service level (network like Unilever).

In the graph we see the curve of the logistic cost and service level. If 5 (1,2 week but we use 5 days for the example) days is the maximum order cycle time, we have to cancel all the zone with more than 5 days, even if include the minimum (from the cost point of view). The minimum of the logistic cost change. We use the point that is in the limit of the allowed area.

Principles 1.2 The number of nodes at level one of the networks depends on a tradeoff between inventory related costs and transportation cost (main trade off In term of cost) subject to service level constrains.

2-echelons network vs 1-echelon network We are considering 1-echelon network as the base option. Now we have already the level one in our network, the potential functions that we can add with the second level are: •

Further reduction in the order cycle time If we have two types of network, one with 4 warehouses at the first level and another with one warehouse at the first level and 4 warehouses at the second level. The capillarity of the network is the same à we are not reducing the order cycle time. This potential function is true only when we have a narrow first level and we have a narrow fist level when we want to keep low the inventory cost. à The benefit is high when the order cycle time target is very challenging + level one is narrow

•

Reduce secondary or last mile transportation costs If we have one warehouse in the north of Italy (MI) and we have to serve customer in Venice province, we go there with a truck visit the customers and go back. If we have many customers to visit, we cannot do it only in one travel, for this reason we use a second level of warehouse in order to be more close to the customer and in this way reduce the travel time and consequently increasing the time for the visiting phase. We use the second level to split the service in the replenishment of the first level for the second level and the visiting of the second level to the customer.

àThe benefit is high when you have many customers ordering small quantities frequently, when your distribution problem is very complex in the last mile and when secondary transportation costs are and important cost component over the overall logistic cost

Examples 2 – Ferrero, Iveco - 2echelons network Distribution problem in Italy Points of origin

Product

Points of destination

Service level

-number -location -specialization -product range -value density -risk of obsolescence

Ferrero -10 -ITA+Eu -focused factories -500,1000 -medium high -very high (perishability)

-number -location -order size -order cycle time

-100000 -ITA -cartoons 1-2 days

Iveco (spare parts) -1000 -world -focused -300000 -very high -very high (goods in inventory reduce value very quickly, difficult decision in defining the parts to have in the point) -100000 -EU -single items Some hours

2-echelon network, first level with one warehouse and second level with 15 warehouses

2-echelon network with one warehouse at level one and 4 warehouses in the level 2

Typologies of network

Why did they choose 2-echelon network? Potential functions

Further reduction order cycle time

Reduction of the secondary transportation cost

Ferrero The order cycle time Is strict for both the company (1,2 days), very difficult to achieve this target with the only first level. But why the first level is so narrow? In the description of the product we see that a high value of product, for this reason we have to concentrate the inventory à narrow first level IMPORTANT The company has a lot of small customer (all the café and supermarket in Italy that order frequently during the week. IMPORTANT

Iveco IMPORTANT (for the same reason)

IMPORTANT (same reason)

It’s clear that they need a second echelon of warehouses. Why Ferrero network is wider than Iveco (very narrow)? Why this difference in term of number of warehouses? For Iveco is more important take the inventory cost under control more than Ferrero because the density value is higher.

Example 3 – Artsana – 2echelons network Distribution problem in Italy Point of origin

-Number -Location -Specialization

Products

-Product range -Density of value -Risk of obsolescence

Point of destination

-Number -Location -Order size -Order cycle time

Service Type of network

The point of origin are manuf. plants

Artsana -10-100 -EU, far east -focused -5000+fashion products - high -medium high because there are a lot of fashion products

Shops, retailers, pharmacies

-30000-40000 -ITA -cartoons -4,6 days -There is one main warehouse and a second level with a transit pointà 2 echelons network with one warehouse at first level and 15 transit point at level two It’s very similar to the network of Ferrero but instead having 15 warehouses they have 15 transit points

Difference with IVECO and Ferrero: At the second level we don’t have inventories because there is only transit point à there is only intransit goods, only for efficiency in transportation and not in the replenishment Potential functions

Artsasa IMPORTANT (even if less than the other but pharmacies order very frequently goods)

Further reduction order cycle time

Reduction of the secondary transportation cost

This is the main difference with the other companies, here we have 4 6 days that is not so strict. Could be achieve with one level. NOT IMPORTANT For this reason, they decide to use transit point and not warehouses

Principle 2 Level 2 has two main potential function • Reducing the order cycle t...