Transcript the PPT file of this seminar

CUSTOMER NEEDS
ELICITATION FOR
PRODUCT
CUSTOMIZATION
Yue Wang
Advisor: Prof. Tseng
Advanced Manufacturing Institute
Hong Kong University of Science and Technology
Advanced
Manufacturing
Institute
Background
Axiomatic design:
Customer
Needs
Functional
Requirements
Design
Parameters
Process
Variables
(CNs)
(FRs)
(DPs)
(PVs)
Product
specification
definition
Product
design
Process
design
CNs are expressed in explicit product specifications.
2
Advanced
Manufacturing
Institute
Introduction

Customer needs elicitation should be
 Good: predictive, customer insight
 Fast: for customers and for designers
 Cheap: reduce market research cost
 Easy: reduce drudgery and errors
3
Advanced
Manufacturing
Institute
Research issues

Can we find what people want quickly and
inexpensively?

How to avoid confusing customers with too many
products?
4
Advanced
Manufacturing
Institute
Challenges

Customers are
 Impatient to specify a long list of items
 Unable to articulate their needs
 Unaware of latent needs
 Lack of information about available options

Interlocking among attributes
5
Advanced
Manufacturing
Institute
Approach

Research framework
 Bayesian network based preferences representation
 Adaptive specification definition procedure
 Recommendation for customized product
6
Advanced
Manufacturing
Institute
Preferences Representation
 Uncertainty of the purchasing choices
 Customers are heterogeneous
 Choice decisions differ under various situations
 The context of purchase differs
 Dependency among preferences towards different
attributes
7
Advanced
Manufacturing
Institute
Preferences Representation
 Bayesian network
8
Advanced
Manufacturing
Institute
Specification Definition

The important considerations in this phase:
 Customers are not patient enough to specify a long
list of items.
 The items differ a lot in terms of the amount of
information they can provide.
9
Advanced
Manufacturing
Institute
Specification Definition

Basic ideas:
 Present the most informative query item to
customers

The value of information:
 f ( X | Y  y) :the additional information received about
X from getting the value of Y=y.
 EY  f ( X | Y  y)  H ( X )  H ( X | Y )
H ( X )   pi log pi
i
10
Advanced
Manufacturing
Institute
Specification Definition

The solution for f (Blachman, 1968#):
EY  f ( X | Y  y)  H ( X )  H ( X | Y )
1
1
f ( X | Y  y )  H ( X )  H ( X | Y  y )   p( x) log
 p( x | y ) log
p( x) x
p( x | y )
x
Y*  arg max EY  f ( X | Y  y )
Y
N. M. Blachman, “The amount of information that y gives about X,” IEEE Trans. Inform.
Theory, vol. IT-14, no. 1, pp. 27-31, Jan. 1968
#
11
Advanced
Manufacturing
Institute
Recommendation

Given:
 Customers preferences information

Determine:
 Which products should be recommended?
 In what order to present the recommendations if
more than one recommendations are presented?
12
Fi ,m ( x)  P X i  x | m 
Advanced
Manufacturing
Institute
Probabilistic relevance computation

Probability of relevance under binary independent
assumption:
P( R  1 | S , C ) 
P(ai | R  1, S )
P(C | R  1, S )

P(C | R  0, S )
i P ( ai | R  0, S )
P( R  1 | S , C )   log
i

pi (1  pi )
 ai
qi (1  qi )
Probability of relevance considering first order conditional
dependency:
P(ai | a (i ) , R  1, S )
P(C | R  1, S )
P( R  1 | S , C ) 

P(C | R  0, S )
i P ( ai | a ( i ) , R  0, S )


p (1  pi )
(1  pi )(1  ri )
p r (1  pi )(1  ri )
P( R  1 | S , c)    log i
 ai  log
 a (i )  log i i
 ai a (i ) 
qi (1  qi )
(1  qi )(1  ti )
qi ti (1  qi )(1  ti )
i 

13
Advanced
Manufacturing
Institute
Probability ranking principle

The idea is to rank products by their estimated
probability of relevance with respect to the information
obtained.

Probability ranking principle is optimal, in the sense that
it minimizes the expected loss.
14
Advanced
Manufacturing
Institute
Schematic framework
Customer
Product development team
Start
Knowledge base
Select suitable
model
Generate the most
informative query
Specify the item
Update knowledge
base
Present
recommendation
Satisfied with the
recommendation?
Y
Confirm the
specifications
N
Configuration
database
Low prob. to find the
feasible configuration?
Provide tailored product
N
Update configuration
database
Process flow
Information flow
End product
15

P
n
i 1
pi
n
Advanced
Manufacturing
Institute
Evaluation metrics


Precision rate

P
n
i 1
pi
n
Recall rate
R

n
i 1
pi
min( n, m)
16
Advanced
Manufacturing
Institute
Evaluation results

The recommendation based on probability ranking can
guarantee the highest precision and recall rate.

If customers’ preferences to all the components are
independent and the potential preferences towards all
the alternatives of an attribute are random, the
specification definition method based on the information
gain has the highest precision and recall rate.
17
Advanced
Manufacturing
Institute
Evaluation results
Parameters setting
Result (# of experiments in which the
precision and recall rate are highest/ total # of
experiments)
m ~ Uniform (3, 13) N ~ Uniform (50, 100)
9,345/10,000
|Ni|~Uniform
m ~ Uniform (5, 15) N ~ Uniform (100, 150)
9,325/10,000
|Ni|~Uniform
m~Uniform (5, 15) N ~ Uniform (1000, 2000)
9,344/10,000
|Ni|~Uniform
m ~ Uniform (5, 15) N ~ Uniform (100, 200)
9,560/10,000
|Ni|~Norm(1, 1)
m ~ Uniform (5, 15) N ~ Uniform (100, 200)
9,603/10,000
|Ni|~Norm(1, 2)
m ~ Uniform (5, 15) N ~ Uniform (100, 200)
9,262/10,000
|Ni|~Norm(1, 0.5)
18
Advanced
Manufacturing
Institute
Evaluation by utility

Preliminaries:
 Fi ,m ( x)  P X i  x | m 
 Stochastically dominate:
If F1,m ( x)  F2,m ( x) , then approach 1 stochastically
dominates approach 2.
19
Advanced
Manufacturing
Institute
Evaluation results

The presented method
 stochastically dominates other approaches.
 is optimal with respect to any nondecreasing utility
function.
20
Advanced
Manufacturing
Institute
Summary




An approach to elicit customers’ preference is
presented.
The model can be used to adaptively improve definition
of product specification for custom product design.
Based on the model, customized query sequence can
be developed to reduce redundant questions.
Product recommendation approach is adopted to further
improve the efficiency of custom product design
21
Advanced
Manufacturing
Institute
Thank you!
Your suggestions & comments are highly appreciated!
22
Advanced
Manufacturing
Institute
Extension to binary independent assumption

Theorem: A probability distribution of tree
dependence Pt(x) is an optimal approximation
to P(x) if and only it’s maximum spanning tree.
[Chow and Liu, 1968]
MST  arg max  I ( xi , x j (i ) )
C
I ( xi , x j ) 
i
P( xi , x j )
 P( x , x ) log P( x ) P( x )
i
xi , x j
j
i
j
Advanced
Manufacturing
Institute
Why customized product design

Well calibrated customized product design can
integrate customers into design activities




Mitigate the side effect of sticky information
Better meet customers’ requirements
Loyalty can be enhanced.
Help identify latent needs guide future product
development
Advanced
Manufacturing
Institute

Lemma 1: Suppose approach 1 proposes n
recommendations in a sequence S1=(r11,r12,…r1n). Each
recommendation r1i has probability p1i to meet the
customer needs. The sequence is arranged such that
P11  P12  ...  P1n . Approach 2 also proposes n
recommendations in a sequence S2=(r21,r22,…r2n).
These n recommendations may be different from the
ones in sequence S1. Similarly, we also have
corresponding probability serial P21  P22  ...  P2n and P1i  P2i
If 1  i  n for all {P2i : 1  i  n}, then X1 stochastically
dominates X2 where Xi is an indicator of the number of
satisfactory recommendations by using approach i.
Advanced
Manufacturing
Institute

Lemma 2: Suppose approach 1 proposes n
recommendations in a sequence
S1=(r11,r12,…r1n). Each recommendation r1i has
probability p1i to meet the customer needs. The
sequence is arranged such that P11  P12  ...  P1n .
Approach 2 also proposes n recommendations
in a sequence S2=(r21,r22,…r2n) which is a
permutation of S1=(r11,r12,…r1n). Then the
distribution of satisfactory product for approach
1 is identical to approach 2.
Advanced
Manufacturing
Institute

Lemma 3: Let U(x) be a nondecreasing utility
function where x is the number of satisfactory
recommendations. Let Xi be an indicator of the
number of satisfactory recommendations by
using approach i. If X1 stochastically dominates
X2, then the expected utility by adopting
approach 1 is greater or equal to that of
approach 2, i.e., .
EU  X 1   EU  X 2 
Advanced
Manufacturing
Institute
Evaluation
m: the number of attributes
ni: the number of alternatives of the ith attribute
m
N: the total number of configurations N   ni
i 1
Pijk: if the jth alternative of the ith component is selected,
the probability that the kth configuration is the desired
one.
The entropy of the configuration space if the jth alternative
N
of the ith component is selected:   pijk log pijk
k 1
The expected entropy of the configuration space if the ith
component is proposed for a customer to specify:
 N

p

p
log
p

ij   ijk
ijk 
j 1
 k 1

nj
28
Advanced
Manufacturing
Institute
Background

Competitive and changing market

Shorter product development time

Product variety proliferation

Bigger penalty cost of failing to meet customers’ needs
or catch up customers’ needs changes
29
Advanced
Manufacturing
Institute
30
Advanced
Manufacturing
Institute
Probabilistic relevance model

Probability of relevance (including first order conditional
dependency):
P(ai | a (i ) , R  1, S )
P(C | R  1, S )
P( R  1 | S , C ) 

P(C | R  0, S )
i P ( ai | a ( i ) , R  0, S )

Parameters setting:
pi  P(ai  1 | a (i )  1, R  1, S )
ti  P(ai  1 | a (i )  1, R  0, S )
qi  P(ai  1 | a (i )  1, R  0, S )
ri  P(ai  1 | a (i )  0, R  0, S )


pi (1  pi )
(1  pi )(1  ri )
pi ri (1  pi )(1  ri )

P( R  1 | S , c)    log
 ai  log
 a (i )  log
 ai a (i ) 
qi (1  qi )
(1  qi )(1  ti )
qi ti (1  qi )(1  ti )
i 

31
Advanced
Manufacturing
Institute

tailor product to different needs

how to avoid confusing customers with too many products
 Can we find what people want quickly and inexpensively
 how to find out if a customer is interested in a virtual which doesn't exist






reducing inconsistent preferences
good: predictive, customer insight: what people buy or how many will people buy
it
fast: for them and for us: it should be fast, doesn't cost so many time
cheap: reduce market research cost: should be cheat
easy reduce drudgery and errors: should be easy for both customers and
designers
That's all the questions in marketing science today.
32