TWO METHODS FORTWO METHODS FORPRODUCT DESIGN SELECTIONPRODUCT DESIGN SELECTION

Shapour Azarm

Department of Mechanical EngineeringUniversity of MarylandCollege Park, MD 20742

Acknowledgements: Hui Li and Cliff Whitcomb

Issues in Product Design Selection

� People involved take on their own �view� of the problem, (e.g., design, marketing)

� Engineer designs a product consisting of multiple subsystems with multiple, conflicting, non-commensurate �performance� objectives

� Customers with different preferences:

� Few customers (e.g., market segments)� Many customers

� Company usually uses profit for measuring benefit

Typically involves multiple aspects, many in conflict:

SELECTION WITH FEW CUSTOMERS*SELECTION WITH FEW CUSTOMERS*

*Whitcomb, C.A., N. Palli, and S. Azarm, �A Prescriptive Production-Distribution Approach for Decision Making in New Product Design,� IEEE Transactions on Systems, Man, and Cybernetics, 29 (3), pp. 336-348, 1999.

Selection Model with Few CustomersSelection Model with Few Customers

Product Development Strategy Management

Production-Distribution Trade Off

iDistribution Model

Production-DistributionCoordination

Product Performance

Sub-submodel

Product CostSub-submodel

Production Model

Market SubmodelCustomer

Distribution Submodel

Company SubmodelCompany

Distribution Submodel

Company Effectiveness

Sub-submodel

Customer Cost

Sub-submodel

Customer Effectiveness

Sub-submodel

Product Model

Production Submodel

Production Model

Design Product

Alternatives MOP(z)

Determine Product

Cost

MOP

From Production-Distribution Coordination

To Customer and CompanyDecision Making Process

To/From Market MOE Information

MOP

Production Model

Production is the optimization of the product performance.

Determine Product Design Attributes

(MOP) Needed for Design

Determine Customer �j�

Product Effectiveness

UP j(X)

Determine Customer �j� Product Life

CostUC j(X)

Determine Customer �j�

Product Effectiveness

UP j(X)

Determine Customer �j� Product Life

CostUC j(X)

Distribution Model

Determine Company

Business Goals (MOE)

Determine Company Profit(NPV of Profit)

To Production-Distribution Coordination

To/From Product MOP Definition

From Production-Distribution Coordination

From Product Design Ouput

From Product Cost Ouput

Determine Customer �i� Effectiveness

(MOE)

Determine Customer �i�

Life Cost (MOE)

NPV(Profit) Benefit (UC i) Benefit (UP i)

Determine Customer Market Information (Market Size, MOE)

BEGIN

Distribution Model

Distribution is the consideration of individuals in the group decision.

Production-Distribution Overview

Production: Multi-Objective Optimization

f1

f2 Pareto Alternatives

Distribution for a customer

U1

U2

Selection Steps with Few Customers

Step 1: Obtain market information

Step 2: Define measures of performance (MOPs)

Step 3: Generate design alternatives, the �best� possible

Step 4: Determine product cost

Step 5: Determine customer and company MOEs and utilities

Step 6: Choose m-best ranked alternatives for customers and company

Step 7: Apply a business strategy to select the alternative

SELECTION WITH MANY CUSTOMERS*SELECTION WITH MANY CUSTOMERS*

*Li, H., and Azarm, S. �Product Design Selection under Uncertainty and with Competitive Advantage,� Transactions of the ASME, Journal of Mechanical Design, 122, pp. 411-418, 2000.

Selection with Many Customers

UNCERTAINTYFEASIBILITY/SUPERIORITY

BUSINESS GOALS

DM�S PREFERENCES

MARKETCUSTOMERS COMPETITION

SINGLE PRODUCT PRODUCT LINE

Overall FlowchartOverall Flowchart

Design Alternative Generation

multi-objective optimization

permutation

levels of attributes permutedPareto solution

Design Alternative Evaluation

designer's expected utility

Monte Carlo simulation:

deterministic business goalsdesigner's utility

sampling of random variablesrepeat

preferences of sample customerscost

competitive productsmarket size,discount rate,cost,life

uncertain/other information:

Approach with Many Customers

Design alternatives:

Di ,i= 1, ..., I

Exogenous variables

Y

Utility

U(NPV, MS)

Choose Di to maximize U

Design alternative attributes:

Ai

DemandRevenue

Cost

Market share:MS

NPV

Start

End

Selection Steps with Many Customers

Step 1: Define the market, attributes, assume distributions

Step 2: Sample customer preferences

Step 3: Generate design alternatives

Step 4: Assume a choice model, obtain demand

Step 5: Estimate NPV and market share of each alternative

Step 6: Construct designer's expected utility function

Step 7: Select alternative with maximum designer utility

Follow ups!

Multiobjective Optimization: Definition

K}1,...,=k 0, = )(h J,1,...,j 0,)(g :{D D : subject to

)}(f..., ),(f),...,({f)( Minimize

kj

mi1

xxxx

xxxxf

=≤=∈

=

f1 (e.g. time/op)

� Pareto/non-dominated frontier

� Preferred optimum solution Preferredsolution

f2(e.g. weight)

•

Feasible domain

Pareto frontier

pg

Market Share and Demand Estimation

Choice model: customer will choose a design alternative with highest utility.

Start

market share for designs

market size

demand for design

alternatives

calculate a customer�s utility for all

design alternatives

compare utilities of all designs and

customer�s current choice to estimate

the customer�s final choice

calculate customer�s utility for all competitive

products

compare utilities to obtain a customer�s

current choice among competitive

products

aggregate customers'

choices

sample size

End

Start

repeat for each customer

1) Compromise: Satisfy all customer markets with a single product.2) Market Capture: Identify most influential customer market and

select their best alternative.3) Market Share: Identify common customer markets and select their

best alternative.

4) Customize Product: Develop the best product for each customer market.

5) Modularize Product: Develop products by varying their characteristics, divided into baseline and optional attributes, to create a line of products that are best for each different customer market.

Business Strategies with Few CustomersBusiness Strategies with Few Customers

Determine Customer �j�

Product Effectiveness

UP j(X)

Determine Customer �j� Product Life

CostUC j(X)

Determine Customer �j�

Product Effectiveness

UP j(X)

Determine Customer �j� Product Life

CostUC j(X)

Determine Customer �i�

Product Effectiveness

(MOE)

Rank Alternatives andIdentify Common Best Ranked Alternatives, Ai

*

SelectBusiness Strategy

Distribution Model Production Model

Determine Customer �i� Product Life

Cost(MOE)

Determine Company Product

DevolpmentBusiness Goals

(MOE)

Determine Customer Market

Information (Market Size, MOE)

Determine Company

Profit(NPV of Profit)

NPV of Profit Outcome

Acceptable?

Select Alternative(s)

Determine Product Design

Attributes (MOP) Needed

for Design

Design Product

Alternatives MOP(z)(MDO)

Determine Product

Cost

Re-adjust Cost for EOS?

Y

NN

Y

END

Compute Distance-to-Ideal

Change Alternative Designs or

Adjust Business Goals or STOP

KeyModeling Information FlowDesign Process Flow

MOP

NPV(Profit) UC i UP i∆∆∆∆z

Ai*

BEGIN

MOP

Approach with Few Customers

Concluding Remarks

Assumptions: Approach with Few Customers

� Segmentation feasibility

� Generation of design (preferably �best�) alternatives

� Stable customer preferences

� No uncertainty

� Known cost, demand, price

Assumptions: Approach with Many Customers

� Generation of design (preferably �best�) alternatives

� Same attributes for generation and evaluation

� New or improved product for an existing market

� Same time-to-market for all alternatives

� First choice model

� Stable customer preferences

� Static market competition

Conclusion

� Selection methods with few/many customers:� Engineering design with marketing aspects� Few or large variety of customer preferences� Competition in market� Implicit demand� Uncertainty

� Directions for future research:� Some of the assumptions need to be relaxed, e.g.,

� Other choice models

� Dynamic market competition