Download Research Methods in Business Fall 2010

Document related concepts

Diffusion of innovations wikipedia , lookup

Transcript
Dr. Stefan Wuyts
Associate Professor Marketing
Koç University
[email protected]
1
1. Social network theory: fundamentals
2. Social networks and new media
3. Social contagion
4. Brand communities
2



Focus on interaction between actors;
Actors (and their actions) are interdependent
instead of independent;
Relational ties are channels for the flow of
resources.
3

The actor
◦ Individual
consumer
◦ Company
◦ Department
◦ SBU
◦ Team
◦ Country
◦ …

Relational tie
◦
◦
◦
◦
◦
◦
◦
Friendship
Transaction
Email
Hierarchy
Biological
Migration
…
4





How do I define ‘actors’ and ‘relational ties’?
What type of variables do I need?
What is the boundary of the network?
How do I sample?
How do I measure?
◦ Unit of observation may not be modeling unit;
◦ Relational quantification
5
 Types of variables



Dyadic structural variables
 Transactions
 Information exchange
 Tie strength
Network structural variables
 Centralization
 Closure or density
Compositional variables (unit of measurement is
the individual actor)
 Firm size
 Personality traits
6
 Boundary
specification: delineating the network
◦ What is your population?
 Sometimes easy, e.g. members of a brand community
 Sometimes difficult, e.g. in networks between
organizations: delineate!
◦ How?
 Based on the perceptions of the actors themselves
(“realist approach”)
 Based on theoretical considerations (“nominalist
approach”)
 Combination of practical feasibility and theoretical
relevance
7
 Sampling:
◦ Random sample draw: Each actor in the
sample is asked about his ties with the other
actors in the sample
◦ Snowball sampling: Actors from the initial
sample are asked about other actors that they
share a tie with; these others are then
included in the sample.
8
◦ Tracing down entire referral chains:
G
Ask actor E which other
actor B recommended
H
F
to consult doctor G,
then contact B and ask
him the same question.
C
A
B
D
E
(Reingen and Kernan 1986)
9
◦ Sampling ego-centered networks: sample networks
consisting of an actor (ego) and the actors s/he is
connected to (alters or neighbors) as well as the ties
among all these actors. Every actor mentions his
partners and suggests what relationships exist
among those partners (oftentimes applied in
anthropology)
10
 Measuring
network data
◦
Unit of observation: often individual actor
◦
Modeling unit: actor (e.g., number of contacts),
dyad (e.g., tie strength), subgroup (e.g., density)
◦
Relational quantification:
Dichotomy (0/1)
Valued
Directional
There is export from
country A to country B
A exports goods worth
1mln to country B
Non
directional
There is trade between
countries A and B
There is intense trade
between countries A
and B
11
 Methods

to collect network data
Survey
◦ Roster vs free recall
◦ Free vs fixed choice
◦ Ratings vs complete
rankings

Archival records

Experimental
Other methods: interviews, observation
See difference with cognitive social structures
(perceptions of others’ ties)
12

Dyad and triad
Dyad = a pair of actors;
Triad = subset of 3 actors;
Step from dyad to triad is fundamental step
A
B
A
B
A
+
-
B
+
C
C
C
Transitivity
Brokerage
Structural
unbalance
13

Network position: centrality
Degree centrality: # direct ties an actor has.
It has been interpreted as the number of information sources available
to the firm (Davis 1991; Freeman 1979, Haunschild 1993)
Betweenness centrality: # times that a focal actor occurs on the shortest
path or geodesic between all pairs of actors in a network
It has been attributed a key role in the distribution of power, as high
betweenness centrality enables the actor to control the flow of resources
to withhold or distort information in transition (Brass 1984; Freeman
1979).
Closeness centrality: average shortest distance between a focal actor and
all other actors in the network.
It has been associated with efficiency in accessing information from the
network (Baldwin, Bedell, and Johnson 1997; Burkhardt and Brass 1990;
Freeman 1979)
Information centr., eigenvector centr. (Google’s PageRank algorithm), …
14

Wasserman & Faust 1997: Core concepts in
social and behavioral theories are quantified
by considering the relations measured among
the actors in a network. (p21)
Let theory dictate your choice!
(Baker and Faulkner 1993)
Beware of the classic problems:
Accuracy – Validity – Reliability - Measurement Error
15

Network structure
◦ Density/closure/social cohesion
A dense network has high degree of transitivity. Basis for trust,
but can be restrictive as well.
◦ Small worlds
“Six degrees of separation” experiments in 60s: claimed that
shortest path length between two Americans was only 6 (on
average).
Puzzling since human social systems are characterized by
transitivity, structural balance, and local clusters.
Watts and Strogatz (1998): small random deviations from
transitivity in combination with local clustering are sufficient
conditions for small world properties
“Small world” properties are observed in many human and
inter-firm networks. E.g., web communities where members
are clustered in friend groups but linkages exist between
friend groups.
16
The strength of weak ties (Granovetter 1973)
Tie strength: two-dimensional construct that refers to the
intensity and valence (positive/negative) of a relationship
Granovetter (1973) found that weak ties are useful for collecting
information and providing access to new networks. Hence:
“weak ties are more instrumental for gathering novel, original
information”.
But Granovetter’s 1973 study was about the search for job leads,
i.e. simple bits of codifiable information.
Weak ties may prohibit transfer of complex and tacit knowledge,
which requires intensive interaction (ability to transfer) and
positive valence (a good understanding  willingness to
transfer). (Hansen 1999)
17
Value of new contact C is function of C’s access to non-redundant networks.
Core is the concept of redundancy (tie strength is merely a correlate, in
some settings)!
18
Brokerage (Burt 1992)
Burt elaborates on redundancy idea:
if actors uniquely connect two parts
of a network they are said to “fill” a
structural hole (cfr. Ruth).
Linking otherwise disconnected parts in the network provides
access to non-redundant information & it provides brokerage
advantage. The broker can exploit its position and only selectively
share information or play other actors against one another.
Brokers have important position in the context of knowledge access
and transfer. Problem: tendency to maintain this unique position
(e.g.: Microsoft as only link between user & software suppliers).
19
Bridging and building (DiMaggio 1992)
DiMaggio points to a more constructive approach of bridging
and building: joining people, ideas, resources from different
networks in a constructive fashion (looking for synergy,
combinational).
Example: thanks to its connections to diverse industries, IDEO
manages to build bridges and solve marketing or design
problems in a particular industry. For example, solutions used
in industry A may be applicable for developing successful
solutions for industry B.
20
Buzz marketing old style: Grab attention, hilarious, taboo,
remarkable: media effort gets leveraged in the press and
in conversations among consumers.
21
22
This is Halfway Oregon,
population 337, which
officially changed it's
name to Half.com) during
the dot-com boom of the
late 90's.
23
Where do new media make a difference?






Connectivity
NPD
Search behavior
Evaluation
Adoption
Loyalty
24
Connectivity
Check who knows whom.
Figure out if you are
really only six steps away
from Kevin Bacon.
25

Involving customers in NPD: crowdsourcing
26

Influence on search behavior: social bookmarking
27

Influence on evaluation: review sites
Share your opinions with others;
Stay ahead of the latest reviews;
Build your own web of trust
28

Influence on adoption: Viral marketing (electronic buzz)
29

Example: Vocalpoint and Tremor (P&G)
Tremor's Women With
Kids Panel
The average mom talks
to 5 people daily. A
connector talks to 20 to
25!
30
31
First studies (o.a. Schmitt, Skiera, and Van den Bulte
2009;Villanueva, Yoo and Hanssens 2008) show
that customers acquired via stimulated word-ofmouth (WOM):
◦ Quit less easily;
◦ Are more effective at generating new WOM;
◦ Have higher contribution margin in short run;
◦ Show more retention in long run.
Not much known yet about return on e-marketing
investments
32

Influence on loyalty: brand communities
33


Social contagion: phenomenon that actors are influenced
in their behavior through exposure to other actors’
knowledge, attitude, or behavior (e.g., imitation).
Causes:
 Awareness and interest
 Satellite dishes
 Impersonal channels dominate
 Belief updating (source credibility)
 Uncertainty reduction, social learning
 Complex products, consumer heterogeneity, source credibility
 Normative pressure
 Theory of reasoned action
 Competitive concerns
 Status
 Complementary network effects
34
Adoption %
100
Laggards (16%)
Late majority (34%)
Early majority (34%)
Early adopters (13.5%)
0
Innovators (2.5%)
Time
35
 Roger’s classification of adopters:
Innovators Entrepreneurial, open to new ideas, higher income
Early
adopters
Opinion leaders, link to early majority, social
networks
Early
Majority
Less leadership, more risk-averse, social networks
Late
majority
Often economic/social pressure to adopt, less
embedded in social networks
Laggards
Not open to change, often adoption after new
versions or substitute products are already entering
the market
36
37
(Source: Tellis, Stremersch and Yin 2003)
38
39
40
Cumulative # adopters
Market potential
dY (t )
 g (t )[ M  Y (t )]
dt
Diffusion speed at
time t
Adoption rate
41
In diffusion model g(t) can take on different
forms:

g(t) = p: diffusion as a function of external
factors. p is called the innovation coefficient.

g(t) = qY(t): internal factors, late adopters learn
from early adopters. q is called the imitation
coefficient.

g(t) = p + qY(t): both external and internal

Bass model: g(t) = p + qY(t)/M
factors.
42
Heterogeneity among adopters: influencers vs. imitators
(Van den Bulte and Joshi 2007)
Some customers are more in touch with new developments and
some (often same) have disproportionate influence on others.
If a proportion θ of the population consists of influentials
(denoted with subscript 1) and the other 1- θ are imitators
(denoted by subscript 2), one needs to account for
heterogeneity in adoption rates:
g1(t) = p1 + q1Y1(t)
g2(t) = p2 + q2[wY1(t) + (1-w)Y2(t)]
(the influentials)
(the imitators)
 Note the asymmetry! Also note: q1 and p2 need not be zero.
 If p2 = 0, contagion from influencers to imitators is critical!
 If p2 = 0 and also w is small, then the diffusion process is
“bimodal”, i.e. the “chasm” pattern (see next sheet).
43
Bimodal diffusion process, with chasm
(p1=0.01; p2=0; q1=0.5; q2=0.2; θ=0.15;
w=0.01)
Adoptions
0.04
0.02
Time
44
Van den Bulte & Joshi 2007: New product diffusion with
influentials and imitators.
Strongly anchored in theory
from both marketing and
sociology
Relatively simple model
extension with major
consequences
Face validity: where model
outperforms alternative models
(and where it doesn’t) makes
sense
Simplicity is also constraining:
two groups only, is that still in
line with middle-status
conformity theory?
No attention to marketing
efforts (despite an earlier
“critical” publication in AJS
(2001) by CvB and Lilien)
Two-step flow hypothesis: information flows to opinion
leaders, and from opinion leaders to opinion followers.
They ideally combine four characteristics:
◦ Interested in and up to date about new products
◦ Being early adopters
◦ Having a central location in the network
◦ Talk about new products
Beware:
◦ No such thing as a generalized opinion leader
◦ Opinion leaders don’t need to combine all four
◦ Misconception that one is either opinion leader or opinion seeker
Also beware of Lake Wobegon syndrome!
46


A brand community is a specialized, nongeographically bound community, based on a
structured set of social relations among admirers of
a brand (Muniz and O’Guinn 2001).
They exhibit:
◦ Shared consciousness of kind (we-ness): the intrinsic
connection that members feel toward one another, and the
collective sense of difference from others not in the
community.
◦ Rituals and traditions: social practices which seek to
celebrate and incalculate certain behavioral norms and
values.
◦ Sense of moral responsibility: felt sense of duty or obligation
to the community as a whole, and to its individual members.
47

Small group brand communities: subgroups in
Harley-Davidson’s “Harley Owners Groups”
◦ Much social interaction with a small group of friends;
◦ Often also beyond the online social interactions;
◦ These small groups are strongly sociocentric (strong
identification with the group as well as the brand),
egalitarian collectivities horizontally organized by
friendship ties;
◦ Group-referent values (social enhancement) are more
important for small group-based virtual communities.
48
49
50

Network-based virtual communities:
◦ Specific group with which the member interacts is less
important;
◦ Participation mainly driven by functional goals;
◦ Limited to online interaction;
◦ Self-referent values (purposive, self-discovery) are
more important for network-based virtual communities.
51
52
Brand
Relationship
quality
-
+
Brand-related
Purchase
behavior
Membership
Continuance
intentions
Community
Membership
duration
Community
Recommendation
intentions
Community
Recommendation
behavior
+
Brand
community
identification
Brand
Loyalty
intentions
Normative
Community
pressure
+
Reactance
-
-
+
+
Community
engagement
+
Community
Participation
intentions
(Arrows without sign are all positive)
Community
Participation
behavior
53
Algesheimer, Dholakia, & Herrmann 2005: The social influence of
brand community: evidence from European car clubs.
Brand communities: natural fit
between social network
literature and CB
Both positive and negative
consequences of brand
communities
Interesting complexity:
community (e.g., size), brand,
customers (e.g., knowledge)
Many hypotheses, several of
which (H5-8) are not
surprising, takes away
attention from the core
Lots of positively valenced
constructs that covary
Reactance is perhaps most
interesting construct, but only
one scale item?
New worry for
marketers?
55