You are on page 1of 15

◆Using Big Data to Improve Customer Experience and

Business Performance
Jeffrey Spiess, Yves T’Joens, Raluca Dragnea, Peter Spencer, and
Laurent Philippart

In an environment where communications service providers (CSPs)


increasingly have the same service offers and devices, offering a superior
customer experience is a priority to compete. Solutions that have the ability
to highlight what really matters in driving customer satisfaction and deliver
actionable insights from their wide-reaching customer, network, and service
data are key differentiators for CSPs. This paper explores ways of integrating
big data insights with automated and assisted processes related to key
customer touchpoints to ultimately improve the customer experience. We
show how innovation from Alcatel-Lucent and Bell Labs helps CSPs improve
their business performance, using unique methodology designed to select
the right key quality indicators, build accurate key business objective
“formula,” predict customer behavior, and ultimately understand which
factors are influencing the most. This can be used for example to improve
the Net Promoter Score (NPS). The net result is a happier customer and a
higher customer value. © 2014 Alcatel-Lucent.

The Customer Experience (3G/4G) smartphones and over-the-top applications.


Communications service providers (CSPs) are Customer experience remains one key differentiator,
facing fierce competition as they strive to win con- when you consider that network reliability, coverage,
sumer and enterprise business with their fixed and care, provisioning, and billing all have an impact on
mobile services. Intense economic pressures, escalat- the customer’s perception of their service provider.
ing consumer demands, and increasingly complex
technologies are raising the stakes, forcing service Measuring Customer Experience
providers to work harder than ever to attract custom- In order to improve customer experience, service
ers and keep them happy. providers must first be able to effectively measure
CSPs can keep customers on board and spur long- and model customer experience. This means that ser-
term success by putting more focus on improving the vice providers must also understand what matters
overall customer experience. The previous areas of most to their customers. Finally they need to make
competitive differentiation such as faster bandwidth, the necessary targeted investments and actions to
unique services, and device innovation have largely optimize customer experience.
disappeared with the advent of fast broadband, multi- Although they hold vast amounts of data about
play services, third generation/fourth generation their network and subscribers, CSPs are not effectively

Bell Labs Technical Journal 18(4), 3–17 (2014) © 2014 Alcatel-Lucent. Published by Wiley Periodicals, Inc.
Published online in Wiley Online Library (wileyonlinelibrary.com) • DOI: 10.1002/bltj.21642
Panel 1. Abbreviations, Acronyms, and Terms
3G—Third generation IT—Information technology
4G—Fourth generation KBO—Key business objective
ACSI—American Customer Satisfaction KPI—Key performance indicator
Institute KQI—Key quality indicator
ARPU—Average revenue per user NPS—Net Promoter Score
BSS—Business support system OLAP—Online analytical processing
CAPEX—Capital expenditure OLTP—Online transactional processing
CDN—Call processing data node OPEX—Operational expenditures
CEM—Customer experience management OSS—Operations support system
CES—Customer effort score QoE—Quality of experience
CLV—Customer lifetime value RoI—Return on investment
CSP—Communications service provider SLA—Service level agreement
DSL—Digital subscriber line SNA—Social network analysis

managing it today. Challenges exist in the fact that (OSS/BSS) reliant on traditional database, data ware-
customer data is “owned” by separate departments, house, and business intelligence tool sets. These tech-
which exist as disengaged silos in the traditional tele- nologies are typically applied to the data in each
com organizations [18]. Each organization has a par- organizational silo, and are configured to create
tial view of its own customer touchpoints, but lacks a reports and dashboards aimed at solving the business
structure in which to piece together the entire end-to- problems of the individual organizations. As the tradi-
end customer view across all of the customers’ touch- tional tools are not scalable and cost effective for very
points with the CSP. Furthermore, the sheer vastness large data sets, very often customer-centric data is left
of customer-centric data that exists across the organi- unanalyzed. Data from multiple organizations in the
zation creates technical challenges in getting to the CSP are not correlated. New technologies designed to
bottom line: what is driving customer experience? handle data on a massive scale have emerged with the
A market segment has emerged for customer experi- buzzword label “big data” technology.
ence management (CEM) analytics tools which prom- Measuring CSP customer experience holistically is
ise to help solve these technical challenges and also to fundamentally a “big data” problem, and to be effec-
overcome organizational barriers [10, 11, 14]. tive, CEM analytics tools must incorporate big data
A good example of a specific measure of cus- technology. Gartner defines big data as having three
tomer experience that is becoming widely used by attributes: high volume, high velocity, and high variety
CSPs today is the Net Promoter Score (NPS*) [16]. [5, 9]. High volume means that there is a growing
NPS is a customer advocacy metric used to rate how quantity of data. This is manifested in the exponential
likely customers are to recommend the product or growth of broadband data traffic, which generates
service. NPS is measured directly and subject to per- multiple billions of customer event records per day in
ception biases and environmental factors. A strong large CSPs. High velocity indicates an acceleration
motivation and business case exists for leveraging in the speed of data, such as the need to detect customer
available customer-centric event data from across the issues in real time to avoid service issues. High variety
different customer touchpoints in order to create an refers to the increase in the types of data, which for the
objective measure of customer experience. CSP brings new challenges for managing data related to
Big Data to the Rescue application usage, web browsing, and social media.
Today CSPs manage their businesses using oper- Big data results in data sets that are so large and
ations support systems/business support systems complex that it becomes difficult for traditional

4 Bell Labs Technical Journal DOI: 10.1002/bltj


databases and business intelligence software to process. advanced analytics algorithms can also help to derive
A number of new technologies have emerged which useful insights and correlation of the high variety
work to solve the challenge of big data. A big data solu- data. For example text and speech sentiment analysis
tion for CEM needs to incorporate these technologies. of unstructured call center transaction logs provides
To solve the high volume challenge, new data insights into the customer’s care experience.
management technologies such as Apache Hadoop*
[3], NoSQL databases such as Apache Cassandra* What Really Matters to the CSP
[2], and next-generation column-oriented data Service providers are in business to maximize
warehouses have been introduced. Hadoop and profit and value to their shareholders. In the end it is
NoSQL technologies like Cassandra achieve scalabil- the money that really matters. However, in order to
ity by adding more and more server clusters (hori- grow and maintain a successful profitable business,
zontal scalability) which process the large data sets in CSPs need to attract and retain customers that are
parallel. Column-oriented data warehouses achieve willing to pay for the products and services they
scalability by organizing the data by columns in the receive. As markets for communications services
relational database, rather than rows, which is much have become hyper-competitive, CSPs have been
more efficient for aggregation calculations over many faced with the choice of either competing on price or
rows but with a limited set of columns. This allows putting more and more focus on customer experi-
parallel access to data across many hard drives rather ence as a way to differentiate and build profitability.
than sequential access across a single drive. Service providers can enhance profitability by
The high velocity challenge needs to be addressed improving the lifetime value of customers. This can
in several ways. First, as customer-centric data is be done with direct actions (such as price increases),
generated, it needs to be rapidly stored. The new data or can be achieved indirectly by managing key busi-
management technologies outlined above each have ness objectives (KBOs) that have an impact on cost
their own mechanisms to accomplish massive fast and revenue. For example, if a service provider focuses
storage, leveraging their parallel architectures. on improving network quality, this can lead to
Customer-centric data must also be processed rapidly decreased customer retention costs and increased
from the data store. Apache Cassandra provides real usage revenue. However this is offset by the costs
time data access, and is very effective in processing required for network capital expenditures (CAPEX)
data on a per-customer basis. Another technology and operational expenditures (OPEX) in order to
known as streaming analytics (also known as event achieve the higher level of network quality. It is a
stream processing), is used to process customer- complex business problem and these complexities
centric data “on the fly” without the need for long- lead to risks; how does the CSP ensure the CAPEX is
term storage. Streaming analytics technology is used spent optimally? Is there a level of quality that achieves
to detect preset conditions and to trigger actions in the best profitability without over-engineering?
real time. For example, streaming analytics can be Service providers have begun to focus on under-
used to detect when a mobile subscriber has experi- standing and optimizing their customers’ experience
enced three dropped calls in the past hour—and can as this has been shown to be a better way to drive up
issue a trigger to a person or an automated system to profitability [17]. By focusing on KBOs that are
take action on behalf of that subscriber. related to the end customer experience, CSPs are able
The high variety challenge can be solved by com- to direct their investments towards improving that
bining the same technologies in a solution architec- experience. By taking a holistic approach and consid-
ture that leverages the technology strengths to make ering the entire customer experience, rather than
sense of the data. Apache Hadoop is useful for storing just the network experience or the care experience,
and processing vast quantities of structured, semi- CSPs can prioritize their spending to tackle the big-
structured, and unstructured data. Domain-specific gest pain points facing their customers. This has

DOI: 10.1002/bltj Bell Labs Technical Journal 5


benefits in creating satisfied customers which on customer effort—whether in the form of solving
average spend more and cost less, with higher cus- customer problems on first contact (first call res-
tomer lifetime value (CLV). olution), ensuring customer effort scores (CES)
are low for key touchpoints, or by being proac-
Customer Centric Management tive with customers to reduce any need at all for
Ensuring that a CSP’s approach to improving the customer to expend effort,
customer experience is appropriate and structured • Removing potential sources of customer dissatis-
first requires an understanding of the primary ele- faction and increasing trust through continuous
ments that constitute customer experience. There improvement—for CSPs this means ensuring
are three areas for a CSP to consider in delivering the networks, services, and operations are meeting
customer experience: their customers functional needs by working
• The customers and what motivates them, flawlessly, and
• The customer perception of products and ser- • Identifying the motivating factors for customers,
vices and their relative importance, and leveraging that knowledge to create moments of
• The market context within which a CSP finds memory or delight and, thus, to generate advo-
itself. cacy which can be as simple as a positive retail
experience and small random acts of kindness as
The work of Frederick Herzberg [6], Noriaki Kano a matter of policy in the care cycle, or as complex
[8] and the American Customer Satisfaction Institute as a seamless multichannel purchase experience.
(ACSI) [12] provide some appropriate frameworks
for understanding these components. Herzberg’s The quest to create sustainable competitive
research, while focused on a worker’s motivation to advantage using customer experience can be summa-
do his job, revealed a fundamental structure for indi- rized as shown in Figure 1.
vidual motivation where factors affecting motivation While the means and modes of interactions
were either creators of dissatisfaction when perceived between people and organizations have changed, the
as inadequate but not necessarily motivating beyond people and cultures they are part of evolve much
being perceived as acceptable (hygiene factors), or more slowly. The understanding of what drives
factors whose presence generated motivation in the human choices as well as their relationships with
individual (motivating factors). products and services are the essential foundations
Noriako Kano’s study and model shows a similar for looking at customer experience in a meaningful
structure for the relationship between customers and way. Today customers interact with CSPs in a multi-
products or services where the absence of essential or tude of ways. These sequences of interactions, or
“must-have” characteristics create dissatisfaction and customer journeys, together form the customer life-
optional or “nice-to-have” characteristics drive satis- cycle: starting with learning about the provider’s
faction. brand, products, and services; deciding to make a
The ACSI’s consistent measurement over time purchase; using products or services; paying for
provides a consistent and methodologically sound them, and at times even having to deal with cus-
proxy for customer experience both within and tomer support. The form of this customer lifecycle is
across market segments within defined geographies often itself a concatenation of different cycles
and markets. between stages of the overall lifecycle.
The implications of these foundations are that For example, the traditional decision making pro-
customer experience measurement could be catego- cess for an intended new purchase is influenced by
rized and focused on three key areas: factors that alter the decision journey [4]. Change in
• Driving behavior to eliminate a perceived defi- the consumer decision journey from the traditional
ciency in hygiene factors by focusing on reducing funnel model (where the consumer considers

6 Bell Labs Technical Journal DOI: 10.1002/bltj


Create irrational
loyalty
Build trust

Remove NPS
customer pain
• Single question
advocacy net-net
Process improvement score drives towards
operational metrics improvement and
• Continuous process and irrational loyalty
CES performance optimization
• Single question VoC
drives towards
optimization of rational
experience for
customer care

CES—Customer effort score


NPS—Net Promoter Score
VoC—Voice of customer

Figure 1.
Creating sustainable competitive advantage.

a number of brands, evaluates them, and makes a Customer Lifecycle


decision to buy followed by focus on the use of The customer lifecycle includes a comprehensive
a product and service) to a more iterative approach set of phases shown in Figure 2 that start with aware-
where the consider/evaluate/buy loop is sometimes ness (user’s perception of the brand), continue with
replaced by the loyalty loop (enjoy/advocate/bond). interact (forming a value perception of products and ser-
This occurs when loyalty to the brand is no longer vices offered and their prices) and are followed by agree/
rational. Therefore, when the consumers’ bond with get (making the decision to acquire a product or ser-
a brand is strong enough, they go directly into repur- vice). The next phase in the customer lifecycle is con-
chase without cycling through the earlier decision- sume which covers the “in-service” experience, followed
journey stages. by support which covers the customer care and payment
Structuring the approach to improving the cus- experience where the simplicity and accessibility of the
tomer experience requires building a bridge between payment interaction is very desirable. The last two
the lifecycle as seen by the customer and its reflec- phases are the reward phase where the CSP offers loy-
tion in the CSP. It means creating an end-to-end alty programs, promotions, and special offers and leave
framework that allows an aggregate as well as a which represents the end of the customer lifecycle.
detailed view of the customer experience at a given Any given customer is less likely to go through
point in time at multiple levels of granularity with the different phases in sequential order than to follow
different frames of reference according to the observ- their own particular journey. For example a customer
er’s span of control. This is a non-trivial task as we sees an ad (awareness), goes online and researches
will show below. the CSP offers (interact) after which he leaves the

DOI: 10.1002/bltj Bell Labs Technical Journal 7


1 2 3 4 5 6 7 8

Awareness Interact Agree/Get Consume Support Pay Reward Leave

Figure 2.
Customer lifecycle.

1 2 3 4 5 6 7 8
Market/ Fulfill Manage Service Charge Retain
Acquire Retire
Sell Order Customer Assurance Bill Optimize

Figure 3.
Communications service provider (CSP) lifecycle.

portal and goes to a shop where he continues the we looked at in the previous section is reflected in a
research (again interact) and decides to purchase CSP journey where the CSP sees the interaction
the product or service (agree). The customer then interrupted on the portal (market/sell) and after a
requests delivery to the home (get). The customer while, sees a person in the shop. The lack of integra-
journey generates “events” along the way, induced tion between systems doesn’t allow for a single view
by the customer interaction with the provider. of the customer. CSPs would benefit from a customer-
CSP Lifecycle centric data model together with a higher level sys-
The CSP lifecycle is a reflection of the customer tem integration. This would allow the service
lifecycle and represents the CSP journey with each provider journey to generate “events” in network
and every customer. Alcatel-Lucent brings to these BSS/OSS, capture process, and network usage activ-
constructs a unique contribution on providing struc- ity and accompany the customer through the life-
tured linkage between the customer lifecycle, the cycle as required.
CSP lifecycle, and the primary business tool that Bridging the Two Lifecycles: How We Measure the
CSPs use to deliver their services. Quality of Experience
The CSP lifecycle follows, as illustrated in Figure 3, The CSP needs to be able to measure and quantify
the customer lifecycle. It starts with acquire, contin- customer experience across all the phases of the cus-
ues with market/sell, followed by fulfill order and man- tomer journey and, if expectations are being met,
age customer. The next phase is service assurance areas which require improvement and areas in which
followed by charge/bill, retain/optimize, and in the end, things are going well. Alcatel-Lucent proposes a model
retire. In reality, an example of the CSP journey is not that defines a holistic view of the quality of experience
a complete list of CSP lifecycle phases but rather a (QoE) by identifying the key drivers of customer satis-
fragmented set of interactions. The customer journey faction across the entire journey. A large component

8 Bell Labs Technical Journal DOI: 10.1002/bltj


of the QoE is related to the “in-service” experience customer journey with the flexibility of aligning the
(or the consume phase) which is connected to the qual- weighting of the QoE drivers to the CSP’s unique pro-
ity of the services and devices used. Often the CSPs file and target and the importance of categories of
focus only on network quality, which is only one part problems that affect the customer experience. The
of the quality puzzle, and do not consider other attri- QoE at the top level should be aligned to the CSP’s
butes such as the quality of the service used, the per- key business objectives (KBOs). These KBOs are in
formance of the device, or its physical design, feature turn calculated from the key quality indicators (KQIs)
set, and ease of operation. The next significant part of and the lower level key performance indicators
the holistic QoE is care which covers the support (KPIs) that drive the KBOs. The holistic QoE approach
experience. Care is linked to the categories of prob- illustrates how big data is used to improve customer
lems that affect the customer experience and the cost experience and loyalty. It allows the CSP to identify
associated with these key drivers. Perception is the what really matters to the end customer. It provides
next QoE component and it maps the awareness, a comprehensive framework that allows the CSP to
interact, and reward phases of the customer lifecycle select the right KPIs, KQIs, and KBOs based on the
to the holistic QoE. This includes the brand image— key drivers to customer satisfaction for an individual
the perception of the value the CSP offers, the loyalty customer, a particular customer segment, or across
programs, and promotions. The remaining customer all customers. Figure 4 shows the holistic QoE in a
journey phases of agree/get and pay are part of the hierarchy of indicators represented by a pyramid.
ease component of QoE. The key drivers of customer The four sides of the pyramid are the key QoE com-
satisfaction related to ease are the billing experience ponents: quality, care, perception and ease. A CSP
and to a lesser degree the activation experience, doesn’t need hundreds of KPIs, only the right KPIs
since there is much more interaction with customers and KQIs to measure and improve customer satisfac-
in the pay touchpoint. tion. Further, an initial selection of metrics is not
The CSP requires an aggregate view of the QoE valid indefinitely. As we explore in the next section,
as well as the view from an individual customer. in a changing environment it is necessary to refine
Depending on the customer’s profile, key drivers to the key drivers behind the QoE using KBO calibra-
their customer satisfaction could be different. For tion and feedback loops.
example a teenager in a family of cell phone users
may be sensitive to the quality of the mobile messag- Actionable Insights
ing service (e.g., text transmission failures and late “Never measure unless you want to record,
text message notifications) while a parent is sensitive never record unless you want to analyze, and never
to the quality of the mobile voice service (e.g., analyze unless you want to take action.” This is a
dropped calls and audio issues) as well as the ease of conventional wisdom that stands the test of time.
using the billing system. Presently, although the data that is available comes
The aggregate view allows the CSP to monitor in more vast amounts, turning it into real informa-
status across all customers as well as by customer tion still requires a touch of art above science.
segment. For example, the CSP might want to iden- In today’s digital networks, the volume of event
tify high value customers to ensure that they enjoy a data is immense, but what kind of insights are CSPs
very high QoE. The CSP would also want to monitor looking for in this data? At large, there are two classes
trends in churn propensity and proactively target of analytics: descriptive and predictive.
retention campaigns towards this customer segment. Descriptive analytics encompass all information
that describes the status or history of the system or
Drilling Down to What Really Matters to the Customer process under investigation. Such descriptive analyt-
The holistic QoE view described in the previous ics can provide the investigator with insights they are
section provides a construct that covers all phases of the required to act upon. Under descriptive analytics we

DOI: 10.1002/bltj Bell Labs Technical Journal 9


1 2 3 4 5 6 7 8

Awareness Interact Agree/Get Consume Support Pay Reward Leave

ty
KBO ali
Qu

Ca
re
KQI

se
ion

Ea
pt
rce
Pe

KPI

KBO—Key business objective


KPI—Key performance indicator
KQI—Key quality indicator

Figure 4.
Customer quality of experience.

find root cause analysis and diagnostics. Diagnostics (DSL) over a period of time may reveal that the root
may involve both the passive reading and interpreta- cause of increased bit error rates can be traced to loop
tion of data, as well as actively triggering particular unbalances or bridged taps or missing splitters.
actions on the system under test and reading out the Figure 5 shows a system where root cause analysis is
results. Root cause analysis is a more elaborate pro- determined subject to Bayesian probability models
cess of iterative digging into data, and correlating [7, 15]. Many more such probability models exist.
various insights such as to determine the one or mul- Predictive analytics take analytics a step further.
tiple fundamental causes of an event. Descriptive Under predictive analytics, data is used to seek to
analytics, although starting from deterministic mea- derive a future state of the system under test, hence
surements, often result in root cause identification to allow for anticipative action. An example here
subject to statistic boundaries. For instance, monitor- would be determination of customers’ propensity to
ing line state information on a Digital Subscriber Line churn, by correlating behavior over a period of time

10 Bell Labs Technical Journal DOI: 10.1002/bltj


Attenuation
55

50

45
OK
40

35

30 NOK Bayesian network


25 Symptom
20
analysis
15

10
Degree
5

22.00 00.00 02.00 04.00 06.00 00.00 10.00 12.00 14.00 16.00 18.00 20.00
Time

Attanution US (3 min) Attanution DS (3 min) Line down


Faiture Collector down

Downstream bit rate


1000
950 OK
900
850
800
750
700 Symptom NOK
650
600 analysis
550
500
450
400
350 Degree
300
250
200
150
100
50
0
22.00 00.00 02.00 04.00 06.00 00.00 10.00 12.00 14.00 16.00 18.00 20.00
Time

Actual bitrats DS (3 min) Attainable bitrats DS (3 min)


ATM traffic DS (5 min) Line down Faiture
Collector down

Data acquisition Extraction of Probabilistic Conclusion


from network significant events reasoning
NOK—Not OK OK—Okay

DOI: 10.1002/bltj
Figure 5.
Line quality diagnosis using Bayesian probability models.

Bell Labs Technical Journal


11
with network event data such as usage records and device management and remote control in today’s
fault indicators. digital services industry.
While analytics may come in various forms, as The second significant driver for the quality com-
stated, the end goal is to influence the system under ponent of the QoE is the quality of the service, deter-
test to correct anomalies, or to further optimize its mined by the network infrastructure, inclusive of
function, changing its behavior. In this paper we communication and application logic. The network
concentrate on those actions that ultimately influ- infrastructure must serve the needs of many custom-
ence the customer experience. ers at the same time, and hence the challenge exists
Different types of analytics may be leveraged for to find a configuration that optimizes the service
use cases such as reactive and proactive care. In reac- offering to many end users. Even more important is
tive care, the CSP intervenes and seeks to correct an ensuring that addressing one customer’s issue doesn’t
anomaly at the time the end user has signaled the result in deteriorated service for other customers. As
degradation or failure of a service. At that point, such, policy control and its impact on optimizing cus-
the elapsed time between the identification of the tomer experience is emerging as a key capability.
issue and the actual resolution is key. Although this This is again where big data analytic capabilities can
includes a long chain of sequential steps, the root “come to the rescue” by influencing network policies
cause analysis in many cases determines the resolu- in such a way as to optimize the overall customer
tion time. In call centers, time is costly, and long wait experience and economies for the CSP.
times may further deteriorate the end-user experi- Complementary to the network infrastructure,
ence. This creates a need for real time or near-real business processes play an important role in the
time analysis mechanisms. overall perceived customer experience. Business
As for proactive care, the CSP seeks to intervene processes that interact with end users can largely be
in the system to anticipate issues and/or to improve on identified as either self-service or assisted-service ori-
the overall customer experience. Here too, one may ented. Both are implemented by business process
seek to influence the behavior of an end user (e.g., logic, which are sometimes called workflows, by
offering free communication time), the end-user ter- which either the end user or a customer service rep-
minal (pushing firmware updates proactively to cor- resentative is guided through a number of steps
rect known deficiencies), the network (e.g., changing interacting with the system.
power levels for mobile terminal communication), or Business processes have more of a hygiene func-
CSP processes (e.g., eliminating a redundant step in a tion in that process anomalies, or missed opportuni-
customer workflow). In the next sections we exam- ties for automation have detrimental effects on the
ine these target systems in further detail. customer experience. By measuring interaction
Earlier we established that the portion of QoE times, and logging events while those processes are
related to the “in-service” experience is connected to being executed, today’s business process systems can
the quality of the services and devices used. The end reveal anomalies like steps that are too complex
terminal is by far the most visible personalization of (exemplified by humans pushing the back button),
the service in the hands of the end user. End user processes that routinely break down before comple-
terminals are always a combination of hardware, tion, or processes that are restarted too often.
software, and customer-specific settings, and are typ- Identifying the actions associated with end users,
ically prone to changes made by the user, either their terminals, and the CSP’s physical assets and
intentionally or unintentionally. Given that the CSP processes is not a unique and one-time activity.
doesn’t have physical access to the device, having the Advanced CSPs are leveraging big data capabilities
capability to remotely perform telemetry and system and analytic tools to create a continuous feedback
changes is key to influence the behavior of the end operation whereby action, measurement, analysis,
terminals, hence the high importance of remote and updated action are repeated on a very frequent

12 Bell Labs Technical Journal DOI: 10.1002/bltj


basis. Advanced CSPs deploy systems that change • Pre-pay recharge. Real time offers to encourage
workflows on a daily basis, leveraging the learning prepaid customers to recharge their account,
experience, even creating target testing groups in depending on their balance.
real time to watch and learn differences about how • Tracking dissatisfied customers. Identify dissatisfied
end-users are interacting with the systems. Modern customers based on explicit feedback (customer
customer experience management solutions should contact), indirect feedback (forum, social media)
then also allow those modifications to occur in an or inferred feedback (usage trends, predicted
agile way in very frequent steps, rather than being QoE) and proactively fix issues or propose attrac-
limited by information technology (IT) release cycles. tive offers at the next customer contact.

Where to Focus? Other high-value use cases include:


Another key aspect of transforming data into • Tracking video experience. Monitor customer expe-
insights, and insights into actions, is selecting the rience when using multi-screen video, identify
customer experience use cases that yield the highest root cause, and optimize video delivery across
financial return for the service provider—those that the network from the source to the end-user
make immediate business sense or are considered device through dynamic line management,
business critical—and starting with the most simple admission control, prioritization, efficient cach-
to articulate with the most demonstrated value. ing, and network/call processing data node
According to a Gartner study [13], an average (CDN) selection.
CSP could potentially generate $10 per subscriber • Predict network outages. Predict network outages
per year in additional margin when using analytics to by analyzing customer complaints and network
turn customer data into insights and subsequently data, and dynamically adjusting network policies
into actions. Six use cases account for 55 percent of to prevent failure.
the total revenue and cost savings: • Track devices with high data usage. Identify devices
• Churn propensity scoring. Uses customer, network, with high data usage to enable efficient quality
usage, and support data to identify key churn driv- of service planning and best fair-usage policy, or
ers and predict the likelihood to churn. Combining optimized tariff plans.
churn propensity with the customer lifetime • Tracking data experience. Monitor mobile data cus-
value allows prioritizing corrective actions on tomer experience at the application layer rather
churn drivers. This is further described in the than the network layer and proactively fix qual-
next section. ity or configuration issues, or propose devices
• Improved self-service. Provide customers with the abil- and tariff plans best adapted to customer usage
ity to diagnose or even control their services from a patterns.
self-care application. Improved self-service includes • Improved assisted care. Use analytics to improve the
proactive alerting, enhanced root cause analysis, efficiency and value of assisted customer care
decision support/recommendation through multi- workflows. Use outage prediction to avoid or
ple channels, and consistency with assisted care. deflect calls. Use customer troubleshooting data to
• Customer segmentation. Segment customers based avoid chronic calls. Monitor customer experience
on usage patterns, value, customer experience, to proactively fix issues and avoid calls. Provide
demographics, or other criteria to improve mar- consistent customer care across multiple channels.
keting, customer understanding, and targeting. • Tracking of enterprise customers. Monitor and report
Used in acquisition, cross-sell/upsell and cus- service level agreements (SLAs) for enterprise
tomer retention campaigns. customers, identify changes in behavior/usage
• Best next offer. Propose personalized offers based on patterns for improved marketing, cross-sell/upsell,
usage, billing, and other customer data analysis. and retention.

DOI: 10.1002/bltj Bell Labs Technical Journal 13


Churn and Customer Lifetime Value per user (ARPU) but may offer lower profitability or
Bell Labs worked with a range of partners to loyalty. Up-selling and cross-selling campaigns are
build an analytics foundation based on developing designed to move subscribers to higher-value CLV
analytic patterns for QoE, propensity to churn, and segments by increasing the spending of loyal custom-
lifetime value. This innovative work combines social ers. Loyalty campaigns are designed to move sub-
network analysis (SNA) and other social factors with scribers to higher-value CLV segments by increasing
more traditional metrics and applies machine-learning their service lifetime. CLV allows network operations
methods to compute the propensity to churn for departments to prioritize the resolution of network
individual users. Bell Labs algorithms serve as the performance issues based on their impact to high-
foundation for proprietary predictive analytics mod- value customers. Over time, a CLV-based segmenta-
els which help CSPs use customer data to identify the tion approach can help increase the proportion of
best opportunities to maximize revenue. customers in higher-value segments.
Use cases based on churn propensity scoring are
at the top of Gartner’s list for potentially generating Addressing the Inhibitors to the Usage of Big Data
additional margin when using analytics to turn cus- In a recent study [1], the top three barriers to
tomer data into insights and actions. Churn predic- implementing customer experience management
tion algorithms foretell the likelihood of a given were difficulty in securing cross-organizational co-
customer to leave a CSP within a given period. operation, poor understanding of the benefits of
Understanding which customers are at risk of churn improving CEM, and poor data quality.
helps the service provider to better target customer To pave the way for a successful CEM strategy to
retention campaigns, resulting in a higher success manage customer experience, service providers must
rate to retain existing customers. By leveraging data empower an executive who is charged with creating
from network and non-network data sources, ana- an enterprise-wide CEM strategy. This executive
lytics provides insights that promote a better under- must control or influence the budget for CEM tools.
standing of how service usage and behavioral With this top-down leadership, the CSP is able to
patterns influence churn. This results in actionable optimize its investment in big data to benefit the
intelligence on key drivers of churn, supporting new entire customer experience. Silo projects and barriers
retention offers that target the root causes of cus- to data sharing are minimized. Service providers
tomer dissatisfaction. have begun to make this change and are appointing
Customer lifetime value (CLV) prediction algo- customer experience officers and others in similar
rithms foretell the profitability of a given customer roles; however in many large organizations the silos
over their service lifetime and enable CSPs to identify are deep rooted, and require a strong management
valuable customer segments to be used in acquisition, focus to overcome. In some countries, market regu-
cross-sell/upsell, and customer retention campaigns. lations also create silos by separating service provider
CLV algorithms help CSPs use customer data to iden- operations into separate businesses such as whole-
tify the best opportunities to maximize revenue. By sale, retail, or infrastructure. A successful CEM strat-
making incremental investments to improve CLV, a egy also needs to consider these constraints.
CSP can improve overall company business perfor- Service providers have begun to develop data
mance and corporate valuation. The concept of CLV management strategies that look to consolidate cus-
can help improve customer segmentation. By build- tomer data across multiple organizations or across
ing a customer segmentation model around clusters the entire enterprise. Software tools for data quality
of CLV, a marketing team can focus on retaining and assurance, data integration, and master data man-
growing high-CLV segments, rather than focusing on agement are used to create a single trusted view of
subscriber groups that generate high average revenue customer-centric data [18]. The immediate benefit

14 Bell Labs Technical Journal DOI: 10.1002/bltj


of this approach is that it maintains “one version of improvements in customer experience, which can
the truth,” so that separate transactional systems and become a barrier for CEM transformation projects.
applications are working with a consistent view of The business consultant helps to estimate cost savings
each customer’s data. Many benefits stem from a from reduced customer churn or more efficient cus-
sound telco data management strategy. Dirty data tomer care, and revenue increases from increased
can be avoided. Once a common repository of cus- personalization and more accurate targeted market-
tomer data is available, it can be leveraged by busi- ing campaigns.
ness analytics to serve the needs of multiple
stakeholder organizations. Conclusion
Data management and consolidation of telco The paper has described how innovation from
data has often been approached using enhanced sub- Alcatel-Lucent and Bell Labs helps CSPs improve
scriber data management products for online transac- their business performance using a unique method-
tional processing (OLTP), and traditional data ology designed to select the right key quality indica-
warehousing solutions for online analytical process- tors, build accurate key business objective “formula,”
ing (OLAP). However, these standard approaches do predict customer behavior, and ultimately under-
not scale to manage the vast scale of customer-centric stand which factors are influencing the most. Alcatel-
telco data, and CSPs have begun to adopt big data Lucent provides structured linkage between the
technology as a cornerstone in their data manage- customer lifecycle, the CSP lifecycle, and the primary
ment strategies. business tool that CSPs use to deliver their services.
It is often helpful for the CSP to bring an external As service providers advance in their implemen-
consultant on board to serve as a trusted advisor to tation of CEM strategies, the direct result will be a
guide in planning the right strategies for data man- more personalized experience for telecommunica-
agement, CEM, and business analytics. The consul- tions customers. With actionable insights and a 360
tant assesses the current state of the CSP versus its degree view of each customer’s behavior, issues, and
peers in both local markets and across the globe, and desires, the CSP provides a better experience by
identifies key gaps in achieving the priority business enabling a positive perception of their company cou-
goals. pled with quality, customer care, and ease of doing
Consultants should be engaged during the entire business that satisfies or delights them. Further
lifecycle of a CEM transformation project. In the research in this area must focus on enabling that pre-
planning phases, the consultant helps with analysis cise personalization that maximizes customer satis-
and benchmarking prior to building the strategy and faction together with customer profitability.
business case. In the implementation phase, consul-
tants help to define the architecture. Following Acknowledgements
deployment, consultants can help by ensuring that The authors would like to acknowledge the fol-
each organization is achieving the maximum busi- lowing members of the Customer Experience
ness value from the newly available tools. Solutions Department for their contribution to this
Technical consultants can help a CSP to design work: David Stevenson, Hilary Mine, Ann Marie
the optimum data management and analytics archi- Vega, Rhodo Odysseos, and Geeta Chauhan.
tecture which incorporates big data technology.
*Trademarks
Business consultants that are focused on customer Apache, Apache Cassandra, and Apache Hadoop are
experience can aid the CSP in understanding where registered trademarks of The Apache Software
to make priority investments that have the highest Foundation.
Net Promoter Score is a trademark and NPS is a regis-
return on investment (RoI). Often it is difficult for a tered trademark of Satmetrix Systems, Inc., Bain &
CSP to understand the business case behind Company, Inc., and Fred Reichheld.

DOI: 10.1002/bltj Bell Labs Technical Journal 15


References [15] J. A. Paulos, “The Mathematics of Changing
[1] Alcatel-Lucent and Heavy Reading, Your Mind,” New York Times (U.S.), Book
“Transforming the Brand Through Improved Review, Aug. 5, 2011, <http://www.nytimes
Customer Experience – Service Provider .com/2011/08/07/books/review/the-theory-
Strategies: Highlights From a Heavy Reading that-would-not-die-by-sharon-bertsch-
Study for Alcatel-Lucent,” 2012, <http:// mcgrayne-book-review.html?_r=1>.
resources.alcatel-lucent.com/?cid=161038>. [16] F. F. Reichheld, “The One Number You Need
[2] Apache Software Foundation, “Cassandra to Grow,” Harvard Bus. Rev., 8:12 (2003),
Wiki,” “Front Page,” <http://wiki.apache.org/ 46–54.
cassandra/>. [17] B. H. Schmitt, Customer Experience
[3] Apache Software Foundation, “Hadoop Wiki,” Management: A Revolutionary Approach to
“Project Description,” <http://wiki.apache. Connecting with Your Customers, John Wiley &
org/hadoop/ProjectDescription>. Sons, Hoboken, NJ, 2003.
[4] D. C. Edelman, “Branding in The Digital Age: [18] M. Sheina and S. Bali, Dialing into Telco Data:
You’re Spending Your Money in All the Wrong Reliable Data Management Drives Smarter
Places,” Harvard Bus. Rev., 88:12 (2010), Telco Business, Ovum, Ref. Code IT014-
62–69. 002697, Feb. 8, 2013.
[5] R. Edjlali, “Getting Value from Big Data,”
Gartner, Webinar, Aug. 14, 2012.
[6] F. I. Herzberg, “One More Time: How Do You (Manuscript approved October 2013)
Motivate Employees?,” Harvard Bus. Rev.,
65:5 (1987), 109–120. JEFFREY SPIESS is director, product management in
[7] E. T. Jaynes, “Bayesian Methods: General Alcatel-Lucent’s Networks and Platforms
Background,” Maximum-Entropy and group in Plano, Texas. He is currently
Bayesian Methods in Applied Statistics (J. H. managing the Motive Analytics product line
Justice, ed.), Cambridge University Press, for customer experience analytics and care
Cambridge, 1986, pp. 1–25. analytics. Since joining Alcatel-Lucent in
[8] N. Kano, N. Seraku, F. Takahashi, and S. Tsuji, 1997, he has held numerous positions
“Attractive Quality and Must-Be Quality” (in managing carrier applications, including software
Japanese), J. Japanese Soc. Quality Control, development, systems engineering, product and
14:2 (1984), 39–48. solution management, product and field marketing,
[9] L. Kart, “High-Tech Tuesday Webinar: Big Data strategy, solutions architecture and consulting. He
Opportunities in Vertical Industries,” Gartner, holds three patents in telecommunications technology.
Webinar, ID No. G00236293, Aug. 8, 2012. Prior to joining Alcatel-Lucent, Mr. Spiess was
[10] S. Kingstone, CEM Vendor Battleground: employed with Texas Instruments where he designed
Insights Lead to Retention and Monetization, hardware and software systems for defense and
Yankee Group, Dec. 2012. telecom applications. He holds a bachelor of science
[11] D. Mendyk and D. Culver, “Service Providers degree in electrical engineering from the University of
Use CEM for Competitive Advantage,” Heavy Cincinnati, Ohio, and a master of science degree in
Reading Insider, 12:5 (2012). computer science engineering from the University of
[12] National Quality Research Center (NQRC), Texas at Arlington.
Stephen M. Ross School of Business,
University of Michigan, American Customer YVES T’JOENS is the vice president and head of Product
Satisfaction Index: Methodology Report, Mar. Development, Customer Experience
2005. Solutions in the Alcatel-Lucent Platforms
[13] C. Patrick, Market Insight: Calculating the group in Antwerp, Belgium. Prior to that,
Value of CSP Customer Data, Gartner, ID No. he served as chief technology officer for
G00239788, Dec. 7, 2012. Alcatel-Lucent’s Access Network Division
[14] C. Patrick, M. Kurth, and M. Cana, Market and as manager of the company’s Digital Home Care
Insight: Improving CSP Customer Experience business. He is the author of multiple papers, holds
with New Monitoring Solutions, Gartner, ID multiple patents, and has actively contributed to the
No. G00215998, Sept. 27, 2011. standardization of access networking.

16 Bell Labs Technical Journal DOI: 10.1002/bltj


RALUCA DRAGNEA is director, product management in LAURENT PHILIPPART is director of the Motive Analytics
the Alcatel-Lucent Platforms group in program in Alcatel-Lucent’s Networks and
Ottawa, Ontario, Canada. She is currently is Platforms Group. This program encompasses
responsible for a suite of content the Alcatel-Lucent Data Management
management products, an integral part of Platform and associated Analytics Templates
customer experience management for fixed, mobile, and cable service
solutions. She has extensive experience in providers. Mr. Philippart has over 20 years
senior technical and management roles within the experience in OSS service assurance and he holds a
communications industry. In her previous role she was telecommunications engineering degree from the
responsible for strategy and incubating new solutions École Nationale Supérieure des Télécommunications de
with a focus on customer experience analytics, Bretagne in France and a master of science in satellite
optimization, security and cloud computing. Prior to communications and spacecraft technology from the
that she was the founder and general manager of the University College London in the United Kingdom.
Web Services Alcatel-Lucent Venture which was From 2006 to 2012 he led the TeleManagement Forum
successfully acquired by the Enterprise Product Group Service Level Agreement Management project and
in 2008. She has served as a Bell Labs research project served as editor of the Forum’s IPTV Application Note.
leader, and has also held director positions in Alcatel- He has been involved in a number of large projects,
Lucent R&D groups responsible for bringing to market including consulting and implementation for
leading products in Internet Protocol (IP), Multiprotocol performance and service quality management systems
Label Switching (MPLS), and Asynchronous Transfer Mode for service providers worldwide and has specialized in
(ATM). Previously she was the principal architect and she voice and audio/video quality assessment for IP-based
founded groups responsible for developing and services. ◆
introducing to market several new and innovative
Newbridge/Alcatel-Lucent products such as 5620
Common Management Information Protocol (CMIP)
Network Management System, 5520 Element
Management System and 5660 Network Design System.
Ms. Dragnea has a master’s degree in electrical
engineering from the University Politehnica of
Bucharest, Romania. She has authored or co-authored
14 papers and holds two patents.

PETER SPENCER is vice president, Customer Experience


Strategy at Alcatel-Lucent. In this role he
leads engagements devoted to helping
Alcatel-Lucent’s customers improve their
own customers’ experience. He has worked
in the communications and information
technology (IT) field for more than 22 years, during
which time he has held executive roles with Alcatel-
Lucent and Ogilvy & Mather. He has a broad
understanding of economics and technologies across
fixed, mobile, and all service types (voice, video and
data). Mr. Spencer’s experience includes customer
experience management and measurement; market
research and operational marketing; brand strategy
and communications management; sales and channel
management, including partner program design and
implementation.

DOI: 10.1002/bltj Bell Labs Technical Journal 17

You might also like