Conference PaperPDF Available

A Social Network Study of the Apple vs. Android Smartphone Battle

Authors:

Abstract and Figures

In this paper we measure and quantify how consumer's choice of smartphones are related to their peers' smartphone choices. Specifically, we study and compare this 'social component' of product adoption for two competing classes of smartphones: iPhone and Android. This is done by constructing a proxy of a social network by using anonymous phone log data from Norwegian mobile phone users, and then coupling adoption data to this social network. We find that smartphone adoption is dependent on the underlying social network both for Android and for iPhone users. Comparing the two, we see that the effect is strongest for the latter. In addition, we measure that the core social network is larger for iPhone users than for Android - Measured by telecom communication, Apple users have more friends than android users. We also present results showing urban/rural differences in smartphone usage.
Content may be subject to copyright.
A Social Network Study of the Apple vs. Android
Smartphone Battle
Johannes Bjelland, Geoffrey Canright, Kenth Engø-
Monsen, Pål Roe Sundsøy
Telenor Research and Future Studies
Telenor ASA
Oslo, Norway
Johannes.bjelland@telenor.com
Rich S. Ling
IT University / Telenor ASA
Copenhagen, Denmark / Oslo, Norway
rili@itu.dk
AbstractIn this paper we measure and quantify how
consumer’s choice of smartphones are related to their peers’
smartphone choices. Specifically, we study and compare this
‘social component’ of product adoption for two competing classes
of smartphones: iPhone and Android. This is done by
constructing a proxy of a social network by using anonymous
phone log data from Norwegian mobile phone users, and then
coupling adoption data to this social network. We find that
smartphone adoption is dependent on the underlying social
network both for Android and for iPhone users. Comparing the
two, we see that the effect is strongest for the latter. In addition,
we measure that the core social network is larger for iPhone
users than for Android Measured by telecom communication,
Apple users have more friends than android users. We also
present results showing urban/rural differences in smartphone
usage.
Social Network Analysis;Viral Marketing; iPhone; Android;
Telecom, product diffusion
I. INTRODUCTION
It has long been known among marketers that our social
network matters when we make purchasing decisions, and that
having positive word of mouth about a product can be a key to
success; see e.g. [1] for a review of studies on social networks
within marketing. Traditionally, data on social networks have
been difficult to collect, but in recent years researchers have
gained access to massive social network data from e.g. online
instant messaging services [8][5] and phone log data
[2][4][3][6][9]. Such data has made it possible to study e.g.
social churn [3], service uptake [2] among telecom customers,
and product adoption on an Instant Messaging network
[8].These studies confirm that consumer behavior is dependent
on the communication network. We have in a recent study [6]
shown how the structure of the adopter networkthe social
network of adoptersdevelops over time, and how social
spreading can be measured by studying this network. In this
paper, we do a comparative study of social spreading effects
for two competing types of smartphones - the Apple iPhone,
and smartphones based on Google’s Android OS.
II. METHOD
Our social network is built by collecting anonymized call data
records, aggregated over a 3-month period, and then using the
communication links (voice and sms) as proxy for the social
relationships. To remove error sources due to non-personal
relationships we have applied some filtering of the dataset.
E.g. we see that some customers have thousands of contacts
during the three months period. This can be machines set up to
automatically send SMSs, company call-centers or other forms
of extreme calling behavior. Such outlier nodes are filtered out
based on combinations of extreme usage and degree (number
of unique contacts). Only traffic between Telenor customers is
used, calls to other operators are excluded. For this study we
have also included weak links we also want to include
relations with limited SMS/voice traffic in the period. In total
we end up with a network containing around 2.5 million nodes
and 45 million edges.
Other studies have shown that mobile phone activity is a good
way to measure real social relationships [5]. We also use
handset type data to associate a handset type with each node in
the social network. With these data we can define the
‘adoption network’ the social network among adopters [6].
This is simply the sub network consisting of adopters and their
common links. We can then study the development of the
adoption network for iPhones (viewed as a single ‘product’)
as seen in [6]and, here, for Android phones, over time
(again making no distinction among the various models of
Android phones). These same data allow us to measure
conditional adoption probabilities between neighbors on the
network, which we use as an indicator of social effects.
Finally, we use postcode informationvery coarse-grained
geographic information on subscribers to map smartphone
adoption to geographical areas in Norway.
III. RESULTS
In a previous paper [6], we looked at the growth of the iPhone
adoption network over time, showing clearly the development
of a ‘social monster’—a giant connected component of the
adoption network which shows the fastest growth. We equated
the strength of this monster with the presence of iPhone
adopters in the ‘dense core’ of highly central subscribers—a
sign of success of the product in taking off. Presence in the
dense core is also inevitably associated with a high density of
adopter-adopter linksa sign that the product adoptions is
‘social’. Here, in using the term ‘social adoption’, we do not
attempt to distinguish homophily effects from true inter-
customer influence: we simply seek to measure the tendency
for those who talk together to adopt together.
In Figure 1, we compare the growth of the Apple adoption
network with that of the Android adoption network, on a
quarterly basis. In each case, we start with the quarter in which
the ‘product’ was first launched. While we see no dramatic
difference in the firstquarter picture (Fig 1(a)), it is clear that
already, two quarters later (Fig 1(c)), the Apple ‘monster’
(Largest Connected Component - LCC) is growing much more
rapidly than the Android monster. This holds not only for total
number of adopters in the LCC, but also in terms of their
percentage of all adopters: two quarters after launch, the
Apple LCC has ca 38% of all adopters, while the Android
LCC has around 28%.
For another indicator of social adoption, we look at the
number of inter-adopter links (adoption pairs) in each
adoption network, over time. Figure 2 tracks the number of
adoption pairs for each product, versus the total number of
adopters. The black dotted curve in Figure 2 gives the number
of adopter pairs expected, for the given total number of
adopter pairs on the fixed call network, if adoption was purely
random. We see that both products generate many times the
number of adopter pairs expected from this random reference
model. Thus, both products show significant social adoption
but, again, the effect is clearly weaker for Android. (The ratio
between the empirical number of adopters, and that number
found in the random reference model, was studied in Ref. [6]
and termed ‘kappa’.)
In Figure 3 we plot yet another indicator of social adoption.
Here we look at
)(kpX
the conditional probability that,
given that a node has k neighbors adopting product X, the
node in question has also adopted product X.
Since random adoption gives a flat
)(kpX
, the positive slope
of the results in Figure 3 are again taken as evidence for social
effects (of some kind) in adoptionfor both products. The
difference between Apple and Android is seen here in that the
Android curve has more weight at small kflattening out at
large kwhile the Apple curve has less weight at small k, but
grows steeply, and almost perfectly linearly, all the way to
k=10. These data were taken in Q3/2011. In this period, the
Apple and Android penetration were approximately equally
(around 18% each). Hence we see that p(k) is
underrepresented at small k (compared to the random case, ie,
a flat line at p(k) = 18%), and overrepresented at large k, for
both productsbut the skew is greater for Apple than for
Android. Taking this skew as an indicator of social adoption,
we find again that Apple is ‘more social’ than Android.
Figure 4 shows the two-dimensional conditional probability
),( YXW kkp
that a node has adopted product W = (Apple or
Figure 1
The figure shows the evolution of the iPhone (Red/Left) and Android (Blue/Right) adoption networks during the 3 first quarters after launch of the first
respective brands. The nodes are customer with iPhone(red) and Android (blue). Links indicate communication between the nodes. Figure a) is the quarter when
the handset first appears in the market, b) is the next quarter and c) is third quarter after product launch. Isolated nodes are not shown i.e iPhone customers that
do not know other iPhone buyers or Android Customers that do not call other Android customers will not appear in this visualization.
Figure 2
The plot shows the number of adoption pairs (Connected customers
adopting same type of handset) vs. the total number of customers having
the brand (x-axis). Red solid line is iPhone, blue stipled line is Android and
black dotted line is the random simulation model.
Android), given that the node has
X
k
neighbors who have
adopted W, and
Y
k
neighbors who have adopted the
competing product. The upper part of Figure 4 shows this for
W = Apple, and the lower part for W = Android. In each case,
the x axis gives the number of neighbors having adopted
product W. Thus we expect (and find) highest weight in the
lower right-hand corner: lots of W neighbors, and few or no
competing neighbors, giving high probability of adopting W.
Hence, qualitatively, neither part of Figure 4 gives a
surpriseand, once again, the observed biasing effects on
neighbors are clearly stronger for Apple than for Android.
Inspired by these results, we have examined the geographic
distributions of these two products. Our method here was to
first aggregate Apple and Android adoption totals over
Norwegian postal codes, and then we take the ratio of the two.
Figure 5 shows the results superimposed on a map of Norway.
What we find, very simply, is that Apple is dominating in
Norway’s cities. Since these results are also from Q3/2011,
there are roughly equal numbers of Apple and Android
phonesso that Apple cannot win everywhere. Thus we see a
rather stark urban/rural dichotomy, with Apple dominating the
cities and Android turning up as scattered blue spots in the
countryside.
We conjecture (but have not yet tested) that the high-centrality
users (as measured by eigenvector centrality) are concentrated
geographically in the cities (just as they are concentrated, by
definition, in the dense core of the social network). In any
case, all of the above results give a picture of Apple users as
being more attracted to other Apple users than are Android
users to other Android usersbut also, more social in general.
To test this idea, we show in Figure 6 the average degree
centrality of three groups: Apple users, Android users, and the
whole population. The result is clear: while Android users
have slightly higher degree centrality than the whole
population, Apple users have around 50% higher degree
centralitythey are more social, by this definition.
Results from an earlier period confirm this picture. We
measured the number of Apple-Apple links and Android-
Android links (as in Figure 2), but also the number of Apple-
Android links. From these data we can get the average number
of Apple and Android friends an Apple user hasand the
same for an Android user. The result was clear: the average
Apple user had over two times as many Apple friends as
statistically expected from no preferencewhile all other
results (number of Apple friends of Android users, and
number of Android friends of Apple and Android users) were
Figure 4
The plots show the adoption probability for iPhone and Android versus
the number of contacts with iPhone and number of contacts with
Android. Top plot: x-axis is the number of iPhone contacts, y-axis is
number of Android contacts and color intensity is iPhone adoption
probability for ego (Dark is low probability).
Lower plot: Color intensity is ego’s Android adoption probability given x
number of Android contacts and y number of iPhone contacts. At the time
of measurement, market share of iPhones roughly equals that of Android
smartphones.
statistically consistent with no Apple/Android preference. In
short: restricted to smartphone users, we again find that Apple
users have more friends, and a stronger preference for their
‘own kind’.
Using our geographic information on subscribers, we have
displayed the results in Figure 6 in terms of three broad
geographic categories—“urban”, “small town”, and “rural”.
Here we see a clear result that is counterintuitive: for all three
groups of nodes (Apple, Android, all), we find that the average
degree centrality increases steadily as one moves from urban
to small town to rural. This result is not understood. We leave
this question (and the measurement of eigenvector centrality
for these groups) to future work.
IV. SUMMARY AND FUTURE WORK
In previous work [6], we showed strong social effects on the
adoption of iPhones. Our aim in this work was to perform
similar measurements on Android phones, in such a way that
we could compare the two types of Smartphones on the
dimension of social adoption effects. We have looked at: (i)
the growth of the social monster in the dense core of the social
network; (ii) the number of adoption pairs; (iii) the conditional
probabilities for adopting Apple or Android, conditioned on
the adoption numbers for adopting neighbors; (iv) the
distribution of Apple and Android over the Norwegian
geography; and (v) the average degree centrality of each
group. Without exception, our results in all these tests support
the conclusion that social adoption effects are considerably
stronger for Apple than for Android. Furthermore, our degree
centrality results imply that Apple users in general are more
socially active than Android users. It seems that Apple appeals
to this type of very social user; and that Apple users in turn
interact strongly, and preferentially, with one another.
Figure 5
The figure shows iPhone and Android uptake mapped to geography via postcodes: Postcodes with more iPhone users than Android users are colored red. Blue
indicates more Android than iPhone in that area. The arrows point at some of the largest cities in Norway. From west to east: Bergen, Stavanger, Kristiansand,
Oslo and Fredrikstad. The visualization shows that iPhone ‘wins’ the cities and Android dominates in rural areas. b) is a close-up of the area around Oslo the
largest city, and c) the area around Bergen the second largest city. iPhone dominates close to the center and in surrounding suburbs.
These results (especially the monster results) imply [6] that
Apple users should have higher (eigenvector) centrality than
Android users. We save this question for future work. We also
plan to look further at the intriguing result, seen here, showing
higher degree centrality in rural areas than in urban areas. We
will also look further into if ‘peer pressure’ effects vary
between urban and more rural locations.
Android handsets are manufactured by several producers, with
varying features. A study on how the social adoption varies
among the different Android brands is reserved for future
work.
[1] C. Van den Bulte and S. Wuyts,"Social Networks and Marketing",
Marketing Science Institue 2007
[2] S. Hill, F.Provost and C. Volinsky, “Network-Based Marketing:
Identifying Likely Adopters via Consumer Networks”, Statistical
Science. 2006, Vol. 21, No. 2, 256-276
[3] Dasgupta, K., Singh, R., Viswanathan, B., Chakraborty, D.,
Mukherjea, S., Nanavati, A. A., and Joshi, A. 2008. “Social ties and
their relevance to churn in mobile telecom networks”. In
Proceedings of the 11th international Conference on Extending
Database Technology: Advances in Database Technology (Nantes,
France, March 25 - 29, 2008). EDBT '08, vol. 261. ACM, New
York, NY, 668-677.
[4] J.-P. Onnela,J. Saramäki,J. Hyvönen, G. Szabó,D. Lazer,K. Kaski,J.
Kertész and A.-L. Barabási,"Structure and tie strengths in mobile
communication networks." Proc Natl Acad Sci U S A. 2007 May 1;
104(18): 73327336.
[5] Sinan Aral, Lev Muchnik, and Arun Sundararajan. "Distinguishing
influence-based contagion from homophily-driven diffusion in
dynamic networks." Proceedings of the National Academy of
Sciences, 106(51):2154421549, December 2009.
[6] P Sundsøy, J Bjelland, G Canright, K Engø-Monsen, R Ling,
“Product adoption networks and their growth in a large mobile
phone network”, IEEE Advanced in Social Network Analysis and
Mining (ASONAM 2010).
[7] N Eagle, A. Pentland, D Lazer, P Alex “Inferring friendshop
network structure by using mobile phone data”, National Academy
of Sciences 106.36 (2009) 15274-15278, 2009.
[8] Rushi Bhatt, Vineet Chaoji, and Rajesh Parekh. 2010. Predicting
product adoption in large-scale social networks. In Proceedings of
the 19th ACM international conference on Information and
knowledge management (CIKM '10). ACM, New York, NY, USA,
1039-1048.
[9] P. Sundsøy, J. Bjelland, K. Engø-Monsen, G. Canright, R. Ling,
"Comparing and visualizing the social spreading of products on a
large-scale social network" , To appear in "The influence on
Technology on Social Network Analysis and Mining, Tanzel Ozyer
et.al (Springer 2012).
Figure 6
Average number of contacts split by customer groups: All, Android users
and Apple Users. The colors indicates location of residency- Urban (grey),
Small town (blue) and Rural (violet).
... Epidemiological models that operate at the con- are not only associated with hardware configurations, but also by their brand. For example, recent intriguing results have found Apple iPhone users to have more connections to others on average, and are more likely to be connected with an iPhone than an Android user [6]. Thus, at the contact level, there may be a higher propensity for information to propagate from one device to another. ...
Article
Full-text available
People in modern society use mobile phones as their primary way to retrieve information and to connect with others across the globe. The kinds of connections these devices support give rise to networks atmany levels, from those among devices connected by near-field radio or bluetooth, to society-wide networks of phone calls made between individuals. This chapter introduces state-of-the-art propagation models that have been applied to understand such networks. It discusses how the models are used in many innovative studies, including how short-lived information spreads between phone callers, how malware spreads within public places, how to detect fraudulent and scamming activity on a phone network, and to predict the propensity of a user to unsubscribe from a mobile phone carrier. It concludes with a discussion of future research opportunities for the study of propagation modeling to mobile phone data analytics.
Article
Full-text available
In the last years, customer churn prediction has been very high on the agenda of telecommunications service providers. Among customers predicted as churners, highly influential customers deserve special attention, since their churns can also trigger churns of their peers. The aim of this study is to find good predictors of churn influence in a mobile service network. To this end, a procedure for determining the weak ground truth on churn influence is presented and used to determine the churn influence of prepaid customers. The determined scores are used to identify good churn-influence predictors among 74 candidate features. The identified predictors are finally used to build a churn-influence-prediction model. The results show that considerably better churn prediction results can be achieved using the proposed model together with the classical churn-prediction-model than by using the classical churn-prediction model alone. Moreover, the successfully predicted churners by the combined approach also have a greater number of churn followers. A successful retention of the predicted churners could greatly affect churn reduction since it could also prevent the churns of these followers.
Book
Full-text available
Propagation, which looks at spreading in complex networks, can be seen from many viewpoints; it is undesirable, or desirable, controllable, the mechanisms generating that propagation can be the topic of interest, but in the end all depends on the setting. This book covers leading research on a wide spectrum of propagation phenomenon and the techniques currently used in its modelling, prediction, analysis and control. Fourteen papers range over topics including epidemic models, models for trust inference, coverage strategies for networks, vehicle flow propagation, bio-inspired routing algorithms, P2P botnet attacks and defences, fault propagation in gene-cellular networks, malware propagation for mobile networks, information propagation in crisis situations, financial contagion in interbank networks, and finally how to maximize the spread of influence in social networks. The compendium will be of interest to researchers, those working in social networking, communications and finance and is aimed at providing a base point for further studies on current research. Above all, by bringing together research from such diverse fields, the book seeks to cross-pollinate ideas, and give the reader a glimpse of the breath of current research.
Article
Full-text available
Studying the use of applications on smart phones is important for developers, handset designers and network operators. We conducted a study on Android devices by installing an instrumentation application, Device Analyzer, on participants’ handsets. Over a 4 month period we collected 10.9 billion records from 674 different users. In this paper we describe how to use the research study features of Device Analyzer to control participant selection and to access information (with consent) that is withheld for privacy reasons from the main dataset. We describe our data processing architecture and the steps required to preformat and analyse the data. Our data contains 3329 distinct applications (from the Google Play store) but despite this, on average, a user makes use of only 8 unique applications in a week. Almost 100% of our users make use of some email application on their phone. Fewer users (85%) made use of the Facebook application but 4–5 times more frequently than for email with sessions lasting almost twice as long. We also investigated whether different applications have correlated usage using a network analysis and a principal component analysis. We see that application usage tends to correlate by vendor more than by activity. This is potentially due to vendors integrating or cross-promoting services between applications.
Article
Full-text available
By combining mobile traffic data and product adoption history from one of the markets of the telecom provider Telenor, we define and measure an adoption network—roughly, the social network among adopters. We study and compare the evolution of this adoption network over time for several products – the iPhone handset, the Doro handset, the iPad 3G and videotelephony. We show how the structure of the adoption network changes over time, and how it can be used to study the social effects of product diffusion. Specifically, we show that the evolution of the Largest Connected Component (LCC) and the size distribution of the other components vary strongly with different products. We also introduce simple tests for quantifying the social spreading effect by comparing actual product diffusion on the network to random based spreading models. As videotelephony is adopted pairwise, we suggest two types of tests: transactional- and node based adoption test. These tests indicate strong social network dependencies in adoption for all products except the Doro handset. People who talk together, are also likely to adopt together. Supporting this, we also find that adoption probability increases with the number of adopting friends for all the products in this study. We believe that the strongest spreading of adoption takes place in the dense core of the underlying network, and gives rise to a dominant LCC in the adoption network, which we call ”the social network monster”. This is supported by measuring the eigenvector centrality of the adopters. We believe that the size of the monster is a good indicator for whether or not a product is going to “take off”.
Article
Full-text available
To understand the diffusive spreading of a product in a telecom network, whether the product is a service, handset, or subscription, it can be very useful to study the structure of the underlying social network. By combining mobile traffic data and product adoption history from one of Telenor’s markets, we can define and measure an adoption network—roughly, the social network of adopters. By studying the time evolution of adoption networks, we can observe how different products diffuses through the network, and measure potential social influence. This paper presents an empirical and comparative study of three adoption networks evolving over time in a large telecom network. We believe that the strongest spreading of adoption takes place in the dense core of the underlying network, and gives rise to a dominant largest connected component (LCC) in the adoption network, which we call “the social network monster”. We believe that the size of the monster is a good indicator for whether or not a product is taking off. We show that the evolution of the LCC, and the size distribution of the other components, vary strongly with different products. The products studied in this article illustrate three distinct cases: that the social network monsters can grow or break down over time, or fail to occur at all. Some of the reasons a product takes off are intrinsic to the product; there are also aspects of the broader social context that can play in. Tentative explanations are offered for these phenomena. Also, we present two statistical tests which give an indication of the strength of the spreading over the social network. We find evidence that the spreading is dependent on the underlying social network, in particular for the early adopters.
Conference Paper
Full-text available
Online social networks offer opportunities to analyze user behavior and social connectivity and leverage resulting insights for effective online advertising. We study the adoption of a paid product by members of a large and well-connected Instant Messenger (IM) network. This product is important to the business and poses unique challenges to advertising due to its low baseline adoption rate. We find that adoption by highly connected individuals is correlated with their social connections (friends) adopting after them. However, there is little evidence of social influence by these high degree individuals. Further, the spread of adoption remains mostly local to first-adopters and their immediate friends. We observe strong evidence of peer pressure wherein future adoption by an individual is more likely if the product has been widely adopted by the individual's friends. Social neighborhoods rich in adoptions also continue to add more new adoptions compared to those neighborhoods that are poor in adoption. Using these insights we build predictive models to identify individuals most suited for two types of marketing campaigns - direct marketing where individuals with highest propensity for future adoption are targeted with suitable ads and social neighborhood marketing which involves messaging to members of the social network who are most effective in using the power of their network to convince their friends to adopt. We identify the most desirable features for predicting future adoption of the PC To Phone product which can in turn be leveraged to effectively promote its adoption. Offline analysis shows that building predictive models for direct marketing and social neighborhood marketing outperforms several widely accepted marketing heuristics. Further, these models are able to effectively combine user features and social features to predict adoption better than using either user features or social features in isolation.
Conference Paper
Full-text available
Social Network Analysis has emerged as a key paradigm in modern sociology, technology, and information sciences. The paradigm stems from the view that the attributes of an individual in a network are less important than their ties (relationships) with other individuals in the network. Exploring the nature and strength of these ties can help understand the structure and dynamics of social networks and explain real-world phenomena, ranging from organizational efficiency to the spread of information and disease. In this paper, we examine the communication patterns of millions of mobile phone users, allowing us to study the underlying social network in a large-scale communication network. Our primary goal is to address the role of social ties in the formation and growth of groups, or communities, in a mobile network. In particular, we study the 'evolution of churners in an operator's network spanning over a period of four months. Our analysis explores the propensity of a subscriber to churn out of a service provider's network depending on the number of ties (friends) that have already churned. Based on our findings, we propose a spreading activation-based technique that predicts potential churners by examining the current set of churners and their underlying social network. The efficiency of the prediction is expressed as a lift curve, which indicates the fraction of all churners that can be caught when a certain fraction of subscribers were contacted.
Article
Full-text available
Node characteristics and behaviors are often correlated with the structure of social networks over time. While evidence of this type of assortative mixing and temporal clustering of behaviors among linked nodes is used to support claims of peer influence and social contagion in networks, homophily may also explain such evidence. Here we develop a dynamic matched sample estimation framework to distinguish influence and homophily effects in dynamic networks, and we apply this framework to a global instant messaging network of 27.4 million users, using data on the day-by-day adoption of a mobile service application and users' longitudinal behavioral, demographic, and geographic data. We find that previous methods overestimate peer influence in product adoption decisions in this network by 300-700%, and that homophily explains >50% of the perceived behavioral contagion. These findings and methods are essential to both our understanding of the mechanisms that drive contagions in networks and our knowledge of how to propagate or combat them in domains as diverse as epidemiology, marketing, development economics, and public health.
Article
Full-text available
Data collected from mobile phones have the potential to provide insight into the relational dynamics of individuals. This paper compares observational data from mobile phones with standard self-report survey data. We find that the information from these two data sources is overlapping but distinct. For example, self-reports of physical proximity deviate from mobile phone records depending on the recency and salience of the interactions. We also demonstrate that it is possible to accurately infer 95% of friendships based on the observational data alone, where friend dyads demonstrate distinctive temporal and spatial patterns in their physical proximity and calling patterns. These behavioral patterns, in turn, allow the prediction of individual-level outcomes such as job satisfaction.
Article
Full-text available
Electronic databases, from phone to e-mails logs, currently provide detailed records of human communication patterns, offering novel avenues to map and explore the structure of social and communication networks. Here we examine the communication patterns of millions of mobile phone users, allowing us to simultaneously study the local and the global structure of a society-wide communication network. We observe a coupling between interaction strengths and the network's local structure, with the counterintuitive consequence that social networks are robust to the removal of the strong ties but fall apart after a phase transition if the weak ties are removed. We show that this coupling significantly slows the diffusion process, resulting in dynamic trapping of information in communities and find that, when it comes to information diffusion, weak and strong ties are both simultaneously ineffective. • complex systems • complex networks • diffusion and spreading • phase transition • social systems
Article
Network-based marketing refers to a collection of marketing techniques that take advantage of links between consumers to increase sales. We concentrate on the consumer networks formed using direct interactions (e.g., communications) between consumers. We survey the diverse literature on such marketing with an emphasis on the statistical methods used and the data to which these methods have been applied. We also provide a discussion of challenges and opportunities for this burgeoning research topic. Our survey highlights a gap in the literature. Because of inadequate data, prior studies have not been able to provide direct, statistical support for the hypothesis that network linkage can directly affect product/service adoption. Using a new data set that represents the adoption of a new telecommunications service, we show very strong support for the hypothesis. Specifically, we show three main results: (1) ``Network neighbors''--those consumers linked to a prior customer--adopt the service at a rate 3--5 times greater than baseline groups selected by the best practices of the firm's marketing team. In addition, analyzing the network allows the firm to acquire new customers who otherwise would have fallen through the cracks, because they would not have been identified based on traditional attributes. (2) Statistical models, built with a very large amount of geographic, demographic and prior purchase data, are significantly and substantially improved by including network information. (3) More detailed network information allows the ranking of the network neighbors so as to permit the selection of small sets of individuals with very high probabilities of adoption.