Inventory Optimization

Inventory Optimization:
Better Inventory Management through Forecasting and Planning

Company INFORM
INFORM Institute for Operations Research and Management INFORM Institute for Operations Research and Management Aachen, Germany Aachen, Germany Advanced Decision Support based on Operations Research, Fuzzy Logic

Optimization

Administrative IT Systems

e.g. SAP and other ERP-Systems, Order Management, etc.

IT Infrastructure Databases / Connectivity

e.g. DB/2, Oracle, Informix, Sybase

e.g. TCP/IP, UNIX, Windows, etc.

First company ever to receive Enterprise Award First company ever to receive Enterprise Award by German Operations Research Society (GOR) by German Operations Research Society (GOR)

Company INFORM
Company Headquaters in Aachen, Germany
250

200

Since 1986, more than 24% Since 1986, more than 24% annual growth each year annual growth each year

150

1h
100

1.5h 2h
50

0 1 2 3 1969 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

2005

Seattle Tokyo Monterrey

Johannesburg Brisbane

Inventory Optimisation Objectives

Item = applicable to Finished Goods Production Parts & Materials Aftermarket Spare Parts Plant Maitenance Spare Parts

Maximising Item Availability

= Maximising Customer Service

100 %

-20 %

Minimising Inventories
0%

Return On Investment

Stock reduction financing IT investment by 100% ...500%

Significant ROI without affecting any labour issues Storage & Capital Costs Discount Pricing Transport & Receiving Planning & Administration

Minimising Logistics Costs

Inventory Optimisation: Essential Functions

Inventory Inventory Controlling Controlling

Demand Demand Forecasting Forecasting

Demand Demand Planning Planning

Capacity & Capacity & Network Network Planning Planning

Inventory Inventory Optimization Optimization

Line Item Line Item Planning Planning

Supplier & Supplier & Shipment Shipment Consolidation Consolidation

Demand Forecasting Matters!

Stock Value

Close to 100 % service level the required stock value tends to 'explode

asting r forc poo

sting od foreca go

High quality forecasting enables High quality forecasting enables to meet any stock availability target to meet any stock availability target with less inventories, with less inventories, and lower logistics costs! and lower logistics costs!

Target Service Level

Factor 1: Forecasting Models

Models & Methods Models & Methods
Simple Moving Averages Simple Moving Averages Exponential Smoothing, e.g. Exponential Smoothing, e.g. Stationary series (SES) Stationary series (SES) Linear trend (DES) Linear trend (DES) Seasonal trend (Holt-Winters) Seasonal trend (Holt-Winters) Dampened trend Dampened trend Irregular demand (Croston) Irregular demand (Croston) Regression Regression ARIMA Box-Jenkins ARIMA / /Box-Jenkins Strong theoeretical Strong theoeretical background but empirical background but empirical results not as impressive results not as impressive Parameter selection not easy Parameter selection not easy Combinations Combinations
No Season No Trend Additive Trend Multiplicative Trend Additive Season Multiplicative Season

Patterns based on Pegels classification for some exponential smoothing methods

Forecasting Model Competitions

Competitions Competitions
Long history Long history M3-Competition M3-Competition [Makridakis, Hibon 2000, IJOF] [Makridakis, Hibon 2000, IJOF] Total of 3,000 different test data sets Total of 3,000 different test data sets 24 methods from many areas 24 methods from many areas Explicit trend Explicit trend ARIMA ARIMA Expert systems Expert systems Neural networks Neural networks Results in nutshell Results in aanutshell

Combinations of models do well Combinations of models do well !! Parameters matter a lot Parameters matter a lot !!
Some M3-Competition results: Average symmetric MAPE for all data
Source: [Makridakis and Hibon 2000]

Factor 2: Forecasting Parameters

Parameter Adaptation Parameter Adaptation

All methods require attention to All methods require attention to 'good' parameters 'good' parameters Hence, forecast quality strongly Hence, forecast quality strongly depends on parameter selection depends on parameter selection Some Examples: Some Examples:
Original data series with noise

Simple moving average forecast with different parameters

DES forecast with better parameters

Auto Adaptive, Dynamic Forecasting

State-of-the-Art Forecasting Models State-of-the-Art Forecasting Models
coping with seasonal factors, trends, coping with seasonal factors, trends, stochastic outliers, and structural breaks stochastic outliers, and structural breaks

Auto Adaptive Forecasting Auto Adaptive Forecasting

Automatic adaptation of model & Automatic adaptation of model & parameters daily weekly parameters daily / /weekly (depending on data volume) (depending on data volume)

Dynamic Forecasting Dynamic Forecasting

Updating the demand profile every time Updating the demand profile every time new data becomes known new data becomes known

Manual Intervention Manual Intervention

Manual modification of forecast where Manual modification of forecast where applicable (e.g. sales promotions) applicable (e.g. sales promotions) Item substitution rules Item substitution rules

Future Demand Profile Future Demand Profile

Inventory Optimisation: Essential Functions

Inventory Inventory Controlling Controlling

Demand Demand Forecasting Forecasting

Demand Demand Planning Planning

Capacity & Capacity & Network Network Planning Planning

Inventory Inventory Optimization Optimization

Line Item Line Item Planning Planning

Supplier & Supplier & Shipment Shipment Consolidation Consolidation

Demand Planning I
key-account structure
r ctu tru
.. .. ..
distribution center

e
..

warehouse

n gio re

s al

a are

customer storage

ion reg

..
y

variant product

ntr ou c

..

product group total product range

rld wo

product structure

Demand Planning II
Multiple Hierarchical Dimensions Multiple Hierarchical Dimensions Permanent Planning Consistency Permanent Planning Consistency: :
All changes will be automatically aggregated and All changes will be automatically aggregated and disaggregated within the entire planning hierarchy disaggregated within the entire planning hierarchy

Alternative use of quantities or values Alternative use of quantities or values Efficient editing of distribution keys Efficient editing of distribution keys

Demand Planning III

Inventory Optimisation: Essential Functions

Inventory Inventory Controlling Controlling

Demand Demand Forecasting Forecasting

Demand Demand Planning Planning

Capacity & Capacity & Network Network Planning Planning

Inventory Inventory Optimization Optimization

Line Item Line Item Planning Planning

Supplier & Supplier & Shipment Shipment Consolidation Consolidation

Line Item Planning I

order quantity X maximum stock

Annual Costs

cumulative costs

reorder point

storage costs total cost minimum ordering + handling costs

Avg. Order Quantity

lead time

Fixed Rythm Order Point Procedures Fixed Rythm / /Order Point Procedures Traditional method Traditional method Simple control mechanism (Kanban) Simple control mechanism (Kanban)

Economic Order Quantity Procedures Economic Order Quantity Procedures Various heuristical algorithms Various heuristical algorithms Optimisation: Wagner Within Algorithm Optimisation: Wagner Within Algorithm Extended to cover discount schemes Extended to cover discount schemes

Line Item Planning II

Ordering Setup Costs Ordering / /Setup Costs Storage Costs Storage Costs Discounts Price Changes Discounts / /Price Changes Supply Calenders Supply Calenders Planned Service Level Planned Service Level Dynamic Safety Stock Dynamic Safety Stock Supplier Contract Specifics Supplier Contract Specifics Minimum Ordering Quantities Minimum Ordering Quantities Packaging Shipment Units Packaging / /Shipment Units etc. etc.

Planning Focus Matrix

Inventory Optimisation: Essential Functions

Inventory Inventory Controlling Controlling

Demand Demand Forecasting Forecasting

Demand Demand Planning Planning

Capacity & Capacity & Network Network Planning Planning

Inventory Inventory Optimization Optimization

Line Item Line Item Planning Planning

Supplier & Supplier & Shipment Shipment Consolidation Consolidation

Supplier & Shipment Consolidation

Order A

Cost optimised ordering schedules

Planning by Line Item

Planning by Supplier

Planning by Shipment

Order B

Inventory Optimisation: Essential Functions

Inventory Inventory Controlling Controlling

Demand Demand Forecasting Forecasting

Demand Demand Planning Planning

Capacity& Capacity& Network Network Planning Planning

Inventory Inventory Optimization Optimization

Line Item Line Item Planning Planning

Supplier & Supplier & Shipment Shipment Consolidation Consolidation

Capacity Planning

Capacity compliant scheduling of production orders No setup time constraints, no order splitting, no sub-task planning Priority based capacity scheduling Identification of bottlenecks, and long-term capacity usage What-if scenario simulation capabilities

Multiple Tier Network Planning

The planning and supply relationship between different plants warehouses is modelled as The planning and supply relationship between different plants / /warehouses is modelled as 'network' for inventory planning purposes. aa'network' for inventory planning purposes.

DC DC DC Plant Regio DC Plant DC DC Global DC Regio DC

DC DC DC

Inventory Optimisation: Essential Functions

Inventory Inventory Controlling Controlling

Demand Demand Forecasting Forecasting

Demand Demand Planning Planning

Capacity & Capacity & Network Network Planning Planning

Inventory Inventory Optimization Optimization

Line Item Line Item Planning Planning

Supplier & Supplier & Shipment Shipment Consolidation Consolidation

Inventory Controlling

Inventory Controlling

IT Support for Inventory Optimisation

ERP // WMS ERP WMS

Focus on data management and information infrastructure very limited math. decision support
(e.g. no auto-adapive parameters)

Supply Chain Supply Chain Management Management

SCM SCM

Numerical Complexity only for key items Decision Making is often de-centralised
(even within one company)

suitable

(100 .. 1,000 <> 10,000 .. 500,000)

Analytical Analytical Systems Systems Statistics, Data Warehouse, Controlling, BI, etc. Focus on 'insight' no direct operations support

Advanced Advanced Planning Systems Planning Systems

APS APS

Math. Models right decision level Coping with large data volumes Day-to-day Operations Support Synchronised with ERP / WMS

Synchronised APS System add*ONE

Download

Automatically released Items

add*ONE Server

add*ONE
Night Processing

Manually released items

ERP
System
add*ONE Clients

Demand Demand Forecasting Forecasting

Line Item Line Item Planning Planning

Item Group Item Group Planning Planning

Capacity/ Capacity/ Network Network Planning Planning

Inventory Inventory Controlling Controlling

... add-on complementing rather than replacing existing IT functions ... add-on complementing rather than replacing existing IT functions

Benefits I
Example: Electrical Tools Manufacturer, Spare Parts Inventory
High service levels despite reduced stock levels. High service levels despite reduced stock levels.
[%] 100 Parts Availability

99

98

[Mio ]

1,900,000.- Inventory Reduction (18%) 1,900,000.- Inventory Reduction (18%)

within months after go-live within 44months after go-live

11

Inventory

10

9 10. 17. 28. 06. 13. 20. 27. 04. 12. 19. 26.

April

May

June

Benefits II
Example: Energy Producer, Plant Maintenance Inventory
Return-on-Investment: Return-on-Investment: 20,000,000.- just months after add*ONE go-live 20,000,000.- just 77months after add*ONE go-live
110,0

105,4

100,0

- 20 Mio

90,0

85,5

80,0

70,0

Start add*One

Start Zentrale Disposition

60,0
M itt el 00 Q 1 Q 2 07 08 09 10 12 11 01 02 03 04 05 06 07 08 09 10 11 12 01 02 03 04 05 08 06 09 07 10 11 12

Benefits III
Return-on-Investment, Example: Small manufacturing company, Return-on-Investment, Example: Small manufacturing company, 30,000 line items spare parts inventory of ca. Mio. value 30,000 line items spare parts inventory of ca. 77Mio. value
3,25 Aufwand (Personen) 3 2,1 2,1

Planning Workload
Staff Hours Reduced by 35%
88 85 80 85

Jahr

Service Level
(Materials & Parts Availability)
Improved by 10%

Servicegrad (%)

6,91 Bestand (Mio. DM)

Jahr

5,39 4,91 4,82

Inventory Level
Reduced by 1.52 Mio. = (22%) in first year; yielding pay-back of invested software within 3 months!

Jahr

add*ONE Selected Reference Clients

Linde Gas

More than 90 companies More than 90 companies in Europe, Asia, South America in Europe, Asia, South America