pipeline inventory

Ford Motor Company’sFord Motor Company’sFinished Vehicle Distribution SystemFinished Vehicle Distribution System

April 2001April 2001

Ellen EwingProject DirectorUPS Logistics

Dr. John Vande VateExec. Director EMILISyE Georgia Tech

ConfidentialConfidentialPage 2

ConfidentialConfidentialPage Page 22

AgendaAgenda

Introduction

1999 Environment

Solution Approach

Network Design

Implement New Strategy

Results to Date

Summary

Introduction



Objectives/MotivationObjectives/Motivation

Importance of Pipeline Inventory

Pipeline Inventory and Network Design

Role of modeling

Information in variables

Stronger formulation

Financial impact



Competitive NecessityCompetitive Necessity

The new BMW Sales and Production System



Financial Incentives: Capital Utilization

– In 1996

– Ford produced 3.9 million vehicles in the US– Avg. transit time 15+ days– Avg. vehicle revenue $18,000– Value of pipeline inventory: > $2.8 Billion– One day reduced transit time:

» $190 Million reduction in pipeline inv.» 1,400 fewer railcars

The Need for SpeedThe Need for Speed




Demand for land• 22 Plants• 54 Destination Ramps• ~1,200 Load lanes• ~8,400 vehicles waiting at plants• $166 Million in inventory




Other Incentives Damage

Flexibility

Others?



Before 1996Before 1996



Origin RampDeparture

Dealer

Arrival at Dealer

InterchangePoint

This diagram represents most of themovements of Ford vehicles.

Copyright 2000. UPS Logistics Inc., All Rights ReservedLast Updated: 8/30/00

FORDTRANSPORTATION

NETWORK OVERVIEW

Destination Ramp Arrival

VehicleForecast

ProductionBegins

VehicleReleased

Origin RampDeparture

Dealer

Arrival at Dealer

InterchangePoint

This diagram represents most of themovements of Ford vehicles.

Copyright 2000. UPS Logistics Inc., All Rights ReservedLast Updated: 8/30/00

FORDTRANSPORTATION

NETWORK OVERVIEW

Destination Ramp Arrival

VehicleForecast

ProductionBegins

VehicleReleased



The PriceThe Price

• Inventory at the cross dock• Added distance traveled• Handling at the cross dock• Capital costs of the cross dock



Mixing CentersMixing Centers

1999 environment



1999 Vehicle Network Delivery Conditions1999 Vehicle Network Delivery Conditions

Record production levels

Demand shift from cars to trucks

Overburdened rail infrastructure

Deteriorating rail service

Shortage of transport capacity

Mixing centers

15+ day transit time

High inventory cost

Dissatisfied customers



High 1999 Level StatisticsHigh 1999 Level Statistics

Assembly plants 22

Mixing centers 5

Destination rail ramps 54

Dealer locations 6,000

Production volume 4.4 Mil./Year

Freight expense $1.5 Bil.

Dec. ‘99 avg. transit time 16.8 Days

Pipeline Inventory $4.1 Bil.



NEVADA

OREGON IDAHO

CALIFORNIA

WASHINGTON

VT. N.H.

R.I.

MASS.

MAINE

MONTANA

WYOMING

UTAH

ARIZONA

NEW MEXICO

COLORADO

NEBRASKA

MINNESOTA

IOWA

KANSAS

TEXAS

OKLAHOMA

MISSOURI

ARKANSAS

TENNESSEE

KENTUCKY

ILLINOIS

WISCONSIN

MICHIGAN

INDIANA

OHIO

GEORGIA

VIRGINIA

FLORIDA

NEW YORK

MD. DEL.

N.J.

HAWAII

W.VA.

MISS.

N. DAKOTA

S. DAKOTA

PENN.

LA.

ALA. S. CAROLINA

N. CAROLINA

CT.

U N I T E D S T A T E S

C A N A D A

M E X I C OGULF OF MEXICO

OCEANATLANTIC

OCEANPACIFIC

Mixing Center Origin Plant Groupings Destination Ramp Planned Ramp Closure

Edison

Norfolk

Atlanta

Kentucky

Ohio

St Louis

CanadaSt Paul

Michigan

Chicago

Kansas City

15% of all vehicles go Haulaway Direct to Dealer within 200-300 Miles of the Assembly Plant

85% of all Vehicles go via Rail to a Hub (Mixing Center or Destination Ramp)

Ford Distribution NetworkFord Distribution Network



Old Delivery DesignOld Delivery Design

Push Network

Vendor sub systems optimized for individual segments

Little to no visibility

Mixing Centers not used effectively

solutionapproach



Ford GoalsFord Goals

Speed 1999: Average 15 days transit time

Goal: Maximum of 8 days transit time

Precision 1998/1999: 37% on time within 1 week

Goal: 95% on time within 1 day

Visibility 100 % Internet vehicle tracking from plant release to dealer

delivery

Guide the flow of vehicles

Respond to variations

Inform customers

network design



Design ProcessDesign Process

Truck vs Rail delivery Allocate Dealers (FIPS)

to Ramps

Route Flows through

Rail Network

Plant

Mixing Center

Origin Ramp

Dest. Ramp

MC Ramp

Ford LocationsFord Locations



An Allocation Model An Allocation Model

Ramps 1 2 3 4 5X 0.8593 0.0862 0.5641 0.8809 0.4220Y 0.8469 0.2032 0.2642 0.3057 0.6329

Dealers X Y Vol.\Cap. 5 6 7 8 11A 0.8384 0.4486 2 0.40 0.79 0.33 0.15 0.46 B 0.1425 0.4539 5 0.82 0.26 0.46 0.75 0.33 C 0.5174 0.6832 5 0.38 0.65 0.42 0.52 0.11 D 0.3865 0.7005 6 0.50 0.58 0.47 0.63 0.08 E 0.8488 0.6002 1 0.25 0.86 0.44 0.30 0.43 F 0.5067 0.1516 4 0.78 0.42 0.13 0.40 0.49 G 0.4725 0.3077 4 0.66 0.40 0.10 0.41 0.33 H 0.5343 0.6397 3 0.39 0.63 0.38 0.48 0.11 I 0.0035 0.4237 2 0.95 0.24 0.58 0.89 0.47 J 0.4080 0.9980 4 0.48 0.86 0.75 0.84 0.37

Total Distance 0

Ramps 1 2 3 4 5

Dealers Vol.\Cap. 5 6 7 8 11Ramps

Assigned

A 2 - B 5 - C 5 - D 6 - E 1 - F 4 - G 4 - H 3 - I 2 - J 4 - Volume Assigned - - - - -

Assignments

Distances

Ford Finished Vehicle DeliveryAllocating Dealers to Ramps

Single Sourcing



Single-Sourcing AllocationSingle-Sourcing Allocation

Var Assign{FIPS, RAMPS} binary;

Minimize TotalCost:

sum{fip in FIPS,ramp in RAMPS} Cost[fip,ramp]*Assign[fip,ramp];

s.t. SingleSource{fip in FIPS}:

sum{ramp in RAMPS}Assign[fip,ramp] = 1;

s.t. ObserveCapacity{ramp in RAMPS}:

sum{fip in FIPS} Volume[fip]*Assign[fip,ramp]

<= Capacity[ramp];



Dealers sourced by

multiple ramps

Old Ramp AllocationSouthern USOld Ramp AllocationSouthern US



Dealers sourced by

single ramps

New Ramp AllocationSouthern USNew Ramp AllocationSouthern US



New Allocation of Dealers to RampsMainland USNew Allocation of Dealers to RampsMainland US



Flows through the Rail NetworkFlows through the Rail Network

Objective is NOT Freight cost!



The Objective ISThe Objective IS

Speed

Capital

Land



The PromiseThe Promise

Speed

Unit trains bypass hump yards




Capital & Land

Time

Laurel, Montana

Orilla, Washington




Capital & Land• 22 Plants• 54 Destination Ramps• ~1,200 Load lanes• ~8,400 vehicles waiting at plants• $166 Million in inventory

Each Plant to One Mixing Center• ~22 Load lanes• ~154 vehicles waiting at plants• ~$3 Million in inventory



The PriceThe Price

• Inventory at the cross dock

• Handling at the cross dock

• Capital costs of the cross dock

• Added distance traveled



Making the Trade-offsMaking the Trade-offs

Measuring Inventory

In the rail network

At the plants and Cross Docks Load-driven system

Railcars depart when full Relationship between

Network Design and Inventory



Inventory at the PlantsInventory at the Plants

Half a rail car full for each destination

Time

Laurel, Montana

Orilla, Washington



Inventory at the Mixing CentersInventory at the Mixing Centers

Half a rail car full for each destination

Time

Laurel, Montana

Orilla, Washington



Workload at the Mixing CentersWorkload at the Mixing Centers

UnpredictableRail car holds 5 vehicles

OrillaBeniciaMira LomaLaurelDenver

Orilla Benicia Mira L. Laurel Denver



Workload at the Mixing CentersWorkload at the Mixing Centers

Balanced: Only load cars you emptyRail car holds 5 vehicles

OrillaBeniciaMira LomaLaurelDenver

Orilla Benicia Mira L. Laurel DenverOrilla



Effect on InventoryEffect on Inventory

Inventory at Mixing Center slowly grows

to just over (ramps -1)(capacity -1) and

remains there Roughly twice the inventory of before Still depends on the number of ramps the

cross dock serves



Consolidation for SpeedConsolidation for Speed

Unit Trains of 15-20 rail cars don’t stop at

mixing yards Trade moving inventory for stationary

inventory



ModelModel

Paths Route from Plant to Ramp Mode used on each edge

Demand[ramp, plant] Combined demand at ramp for all products from the plant

Variables: PathFlow[path]:

Volume from the plant to the ramp on the path

UseLane[fromloc, toloc, mode] binary Did we use the mode between two locations



ModelModel

Objective Minimize the number of vehicles in the pipeline

Moving Component (Transit times) Waiting Component (Mode Size)

Minimize PipelineInventory: sum{path in Paths} (Total Transit Time)*PathFlow[path]; sum{(f,t,m)} (Size[m]/2)*UseLane[f,t,m]



ModelModel

Satisfy Demand The sum of flows on all paths between a plant and a

ramp must meet demand s.t. SatisfyDemand[p in PLANTS, r in RAMPS}:

sum{path in PATHS: Plant[path]=p and Ramp[path] = r} PathFlow[path] >= Demand[p,r];



ModelModel

Define UseLane For each pair of locations and mode between them write a

constraint for each plant and ramp s.t. DefineUseLane[p in PLANTS, r in RAMPS, (f,t,m) in EDGES}:

sum{path in PATHS: Plant[path]=p and Ramp[path] = r and (f,t,m) in PATHEDGES[path]} PathFlow[path] <= Demand[p,r]*UseLane[f,t,m];



ModelModel

Large Model Lots of Variables: Many Paths Lots of Constraints: DefineUseLane

The LP relaxation is nearly always integralUse Column Generation



Reduced plant

destinations

New Rail LanesNew Rail Lanes



NEVADA

OREGON IDAHO

CALIFORNIA

WASHINGTON

VT. N.H.

R.I.

MASS.

MAINE

MONTANA

WYOMING

UTAH

ARIZONA

NEW MEXICO

COLORADO

NEBRASKA

MINNESOTA

IOWA

KANSAS

TEXAS

OKLAHOMA

MISSOURI

ARKANSAS

TENNESSEE

KENTUCKY

ILLINOIS

WISCONSIN

MICHIGAN

INDIANA

OHIO

GEORGIA

VIRGINIA

FLORIDA

NEW YORK

MD. DEL.

N.J.

HAWAII

W.VA.

MISS.

N. DAKOTA

S. DAKOTA

PENN.

LA.

ALA. S. CAROLINA

N. CAROLINA

CT.

U N I T E D S T A T E S

C A N A D A

M E X I C OGULF OF MEXICO

OCEANATLANTIC

OCEANPACIFIC

Mixing Centers Destination Ramps

Union Pacific CSXT FEC

BNSF Canadian Pacific Car Haul to Ramp

Norfolk Southern Canadian National

Edison

Norfolk

Atlanta

Kentucky

Ohio

St Louis

Canada

St Paul

Michigan

Chicago

Kansas City

Final Outbound Rail Network with CarriersFinal Outbound Rail Network with Carriers

resultsto date



ResultsResults

Cut vehicle transit time by 26% or 4 days

$1 billion savings in vehicle inventory

$125 million savings in inventory carrying costs

Avoid bottlenecks

Reduce assets in supply chain

Improved inventory turns at dealer



Benefits Benefits

Ford

Dealers

Rail Carriers Auto Haulers



Benefits - FordBenefits - Ford

On-time delivery

Competitive edge

Cost control



Benefits - DealersBenefits - Dealers

Reduced inventories Increased customer satisfaction



Benefits - Rail CarriersBenefits - Rail Carriers

Improved equipment utilization (reduced capital expenditures) Visibility and planning capabilities Synergies with existing UPS traffic Increased cooperation



Benefits - Auto HaulersBenefits - Auto Haulers

Expanded dealer delivery hours Visibility and planning capability Improved asset utilization Increased cooperation



Other OpportunitiesOther Opportunities

Where to go from here?

pipeline inventory

Documents

ramp mc ramp

destination ramps

destination rail ramps

rail network

days pipeline inventory

days transit time goal

days transit time precision

visibility mixing centers