Internet futures:
growing market intelligence to allow telecoms operators to
seize the opportunities of the information age
Frederic Lagacherie
BT Laboratories
Martlesham Heath, ADMIN2 PP7
Ipswich, IP5 3RE, UK
+44 - 1473 - 642397
frederic.lagacherie@bt.com
Abstract
With the introduction of new access technologies in carrier networks, as well as the
expected expansion of the bandwidth to the end user, telecoms operators are
triggering extensive business analysis to understand and hopefully anticipate the
evolution in total bandwidth demand. In the face of great uncertainty in peak IP
traffic forecasts and the need for significant investments in network facilities, it is
vital for decision-makers to have the tools that will help to manage the risks related to
these investments.
The challenge for modellers here is to conceive the Internet as a complex system of
convergence where the combination of humans, technology and their interactions are
responsible for the future development path. In that respect, we have taken a socio-
technical perspective to break the perceived complexity down to a mode where
behaviours can be simulated dynamically over time. The objective of this model is to
estimate bandwidth demand by sizing the Internet market evolution in terms of the
numbers and types of subscribers.
This paper[1] details the soft-variables and relationships that have been used within a
System Dynamics model to create a seamless environment, in which to explore
potential market scenarios and test various strategies so as to optimise customer
retention and profit from the point of view of an ISP.
Introduction
The oligopolistic environment in which telecoms operators have to compete has
forced them to adopt more complex processes to support their decisions. As
consumer demand asks for more tailorisation and value for money, competition has
lead telecoms operators to take faster strategic decisions including changing the focus
of their core business. The Internet [2], in that respect, offers an unprecedented
wealth of both opportunities and threats for emergent and incumbent actors.
Developing from providing simple access to residential customers, ISPs have emerged
as the brokers between content, online applications, telecoms operators and the end
customer. Since ISPs now own the customer interface, and consequently the user’s
profile, and since traditional telephony applications are being adapted to the IP
paradigm, (voice, fax and call centres to say the least), ISPs could relegate telecoms
incumbents to the unpleasant role of being just bit carriers.
To escape the over-optimistic/pessimistic future drawn from linear forecasting,
telecoms operators must look at random and decision-lead events that may act as
break points in the way the Internet market will evolve. The recent drive by ISPs to
provide Voice over IP (VoIP) services belongs to that range of events that may
threaten the telecoms operators “cash cow”, i.e. the conventional voice telephony
business. However, as humans, we are limited in our ability to mentally shape a model
that can capture the dynamic complexity of the Internet industry and the likely
consequences of our interactions with it. With very few references to history possible,
this paper describes a System Dynamics model that provides a template of product
adoption, competition and corporate decisions for the take-up of Internet services.
The engine behind the model allows the user to assess in real-time how each decision
or change in the environment will affect the way consumers connect and make use of
the services over time. For instance, some Internet service providers were ready three
years ago to launch a free subscription Internet access service. Although this business
model seems to be effective today, it did not appear so at that time. This was mostly
due to the lack of sponsors ready to believe in sufficient return on their investments
through advertising and online purchasing. By simulating Internet demand from a
bottom-up approach, this model will help us to understand the long-term implications
of today’s tactical decisions upon telcos’ network provision business. It will also
emphasis the type of ISP business model that can be sustained in a competitive
marketplace.
Background to the model
In order to focus the scope of this paper on to a manageable number of variables, the
model proposes to consider a competitive, liberalised telecoms market for residential
customers in Western Europe. The end model consists of three key modules, as
shown in figure 1.
Marketing variables
Adoption Module
Network dimensioning Network Module Optimisation
rules Rules
Cost and revenue variables Economic Module
Figure 1. Theoretical structure of the Internet model
This paper will concentrate of the functionality of the Adoption Module. The goal
behind the Adoption Module is not to describe why residential customers behave in a
certain way, but rather to investigate how their behaviour can trigger various scenarios
and also to assess the effectiveness of a policy in influencing these behaviours.
Typically, the Adoption Module creates a set of data that inputs into the Network
Module in order to observe the impact of each scenario on dimensioning of the
network. Data requirements from the Network Module feed in turn into the Economic
Module to monitor the cost effectiveness and revenue optimisation of a given policy.
The approach taken to develop this descriptive model involves initially identifying
the level of complexity that should be put in the design of the model. A first set of
variables has been identified from interviews with professionals and scans through
various internal and external market researches. Some sensitivity analysis was
subsequently performed to identify the influential variables and to remove other
variables that only added to the complexity. In order to further create a realistic
simulation environment, the model has been developed as customer-centric. That is,
the model’s loops and valuation variables (attractiveness, price, performance) have to
be understood as a function of the customer’s perception.
The competitive environment is composed of various actors of all sizes and
geographic scopes, ranging from a local ISP targeting a niche market to the global
OSP. Six of these have been retained for their divergence in setting their strategy,
attracting investments and their respective maturity in this business.
The products described for the scope of this paper are restricted to traditional ISP
services including email, web browsing, ftp, fax and Voice over IP.
The customer segments have been clustered according to a socio-demographic
classification. It has been shown that the Internet is spreading across demographic
layers as a function of revenue and education. Further studies suggest that a sub-
classification should encompass a distinction between family and non-family.
Residential customer segmentation must be considered particularly carefully as
personalisation is becoming the marketing artefact by which companies are seeking to
increase their share of customer [3]. As it will be shown later, the value of an ISP
should not be regarded as the number of its subscribers, but as the frequency at which
subscribers connect as well as their attitude once on-line.
Application of System Dynamics to the adoption of Internet services
The module uses the multiple combinations of product adoption, competition and
corporate decisions to map the complexity of the Internet market. Some of these
aspects will be discussed in this section.
a. Product adoption
The diffusion of Internet products is believed to follow a Bass diffusion model, as
innovation and imitation drives the evolution of today's applications and services that
use the Internet as a transport mechanism. The model will be limited in the number of
connected households by the overall penetration of Internet access technology,
(mainly PCs, TVs and telephones). This limit is materialised in figure 2 below,
extracted from a Powersim model, by the upper blue line.
Total residential market
Penetration rate
Connected household
Figure 2. Total take-up for Internet services in country WE
Although it is possible for one household to own several access devices, the model is
configured to avoid double counts. The lower blue line accounts for the market
inertia, the limit after which ISPs will have captured the easiest part of the potential
customers, (i.e. the innovators, early adopters and early imitators). Once this level of
inertia reached, the total take-up curve (1) for the country considered will behave as if
the market was entering its stage of maturity.
The traditional S curve should, however, be subject to variations inherent to the high
volatility of the Internet market. Considering figure 2, the model shows the
distortions that can be obtained by introducing internal soft-variables, such as
customer experience and awareness or the overall state of the economic cycle, and
their influences on subscriber responsiveness. The purple curve (2) has been
simulated over different environmental conditions and also various price and quality
incentives. These distortions need to be fully understood as the model will translate
the Adoption Module’s outputs into their effects on network load and the subsequent
dimensioning requirements. It is central to this model to be able to show the impact
of various policies and market turbulence on the aggregate take-up of services and
their usage.
As the graph in figure 3 shows, special attention has also been paid to the
responsiveness of the model. From the recent example derived from the introduction
of free subscription Internet access, the model runs over monthly periods with a
maximum delay in the response time for switching ISPs being one or two months,
depending on the nature of the policy stimuli. To make full use of the model's
interactivity, the Adoption Module should be considered to be an automated war game
whereby six players take pro-active decisions to determine the strategy they each
should follow, given the responsiveness and behaviours of rival companies.
The first data displayed in figure 3 represents actual, historical data up to and
including the last quarter of 1998. In this specific scenario, it should be noted that all
companies do react similarly as the market is subject to structural changes and
turbulence.
Household connected per ISP
Figure 3. Cumulative Internet take-up split by ISPs
When individual decisions are enabled from 2001 to 2002, each company behaves in
response to the policy adopted and the subsequent changes in the overall competitive
situation, (in this case caused by a price war). When no new inputs are allowed into
the model the simulation engine simply stretches the output from the last period
linearly, as shown in figure 3 from period 2002 onwards.
b. Competition
From the early years of the Internet, the hype over the potential benefits promised to
those who joined the bandwagon has shaped a very responsive marketplace. Because
most ISPs fear the market consequences of missing even one opportunity, it is
especially hard to differentiate between them, either commercially or technologically,
on a sustainable time scale. The high level of competition watch has given rise to an
extremely rapid catch up rate. This translates into the model partly through the
attractiveness of one company to another and partly via the churn level.
Customer awareness
Product
ATTRACTIVENESS SS a
offer
a,
Promotion
effort
Relative
subscription price
+ connection
fees
Relative
quality of
service
4 Valuation of price VS QoS
Figure 4. Factors contributing to an ISP attractiveness
Within the model, the potential customer base will be split according to each
company’s relative differential attractiveness. In that respect, the attraction factor
encompasses the elements depicted in figure 4 above.
The model will consider the ISP’s respective growth ratio as the mechanism that
triggers “word of mouth” changes in customer base rather than using market share to
do so. This is because, from a customer point of view, choosing an ISP is often the
result of following mass opinion. In that respect, a fast growing company will benefit
from general media coverage, which translates in the model by a boost in its
attractiveness and the resulting customer awareness of that brand. It is also being
suggested that an ISP will hit a large number of its targeted audience by advertising
through traditional media channel, such as television. The promotion effort takes
these parameters into consideration. However, the low churn observed until 1998
amongst the industry has highlighted the lack of comparative information to support
customers’ buying decisions. This suggests that the differential attractiveness factor
should be weighted by the average awareness within each socio-demographic
segment.
Although customer churn could initially be applied as an average monthly percentage
to turn around the lack of commercial data, it would not reflect the high volatility that
characterises today’s Internet marketplace. With number and email portability as well
as the emergence of brokers, it is very convenient and cost effective to swap from one
ISP to another. As a result, the delay between a significant variation in price or QoS
within an ISP offer and the resulting reaction in customer churn will only be of one
period, i.e. a month. An example of this is shown in figure 5. Here, ISP3 modifies its
commercial offer in the first semester 2000, as a result of its bad churn performance
recorded during previous periods. In a competitive marketplace, rival companies are
expected to follow the same tactics on a responsive basis, unless they opt for a
different strategy.
Market churn
Figure 5. Periodic churn rate split up by ISPs
However, if a “gap” between two ISP offers remains for too long, (in this model no
more than two periods), this will be translated into a brutal increase in customer
migration towards the most appealing commercial offer. Here, ISPs 2 and 5 lose
subscribers to ISP 3.
As a result, customer volatility can have a dramatic effect on ISP performances due to
the little delay introduced by the ISP in response to changes in the market [4]. There
is however a longer delay when it comes to switching customer focus from one area to
another, e.g. from price to quality. Finally, when the market reaches a state of
equilibrium beyond 2004, the churn rate behaves as a factor of growth only. This
resilient churn is indeed the result of the constant flow of volatile customers in and out
of ISPs' customer base.
c. Corporate decisions
As we saw in the previous sections of this paper, the take-up of Internet services is
partly a function of price, quality of services (connection failures, connection rate,
customer support, ability to customise billing) and availability (whether the customer
is passed by some access infrastructure, or whether the service is available from one
ISP). The market bears low product differentiation. The ability to attract customers
will be down to the combined used of these variables based on a sensitivity analysis to
point out the primary affecting factors.
Although the above variables are the most influential in the players’ ability to increase
their market share throughout the model simulation, promotion and branding are also
of significant importance. Another way to consider the objective of the simulation,
apart from optimising the total number of customers should be to increase the
customer's respective value for the company, i.e. the share of customer [5]. As long
as the number of new subscribers exceeds customer churn, the balance remains
positive. However, when the market enters its maturity phase, the model emphasis
the importance of retaining customers and increasing their spend throughout their stay
with the company.
Amongst the performance ratios available to monitor the development of one
competitor, i.e. market share, benefits, churn rate, cost per subscribers, see figure 6,
the one that will increasingly attract attention is the customer value to the company.
ISPs would rather benefit from loyal, frequent user subscribers than from volatile
customers who follow the most appealing offer. In the model, this takes the form of a
portal ratio, i.e. the ability of an ISP to grow a more stable customer base. This ratio
is based on whether the ISP provides content services, a search engine mechanism,
news partners, online ticket booking service, etc. The general perception of a brand
and the level of investment in advertising campaigns will also affect this ratio. A high
portal ratio will in turn limit the effect of volatile churn. While the aggregate number
of customers matters when calculating market shares, the split between in-
frequent/frequent users will be vital to calibrate the input to the Network Module so as
to look at usage patterns across a carrier’s network.
Performance.control pannel
baa
Cost per customer Revenue per customer
Portaleratio =
comparison 1 Market
Financialepertormance
TS pee REED oO
iporzs sc00,p00.02) san p00.
Figure 6. Performance control panel for ISP 1
Calibration of the outputs
Customers do not respond to minor changes in policies. That is, if a new ISP enters
the market following a strategy of imitation with no particular perceived utility
benefits to the customer, the market will remain unchanged. As a new entrant, a free
subscription ISP offers sufficient market appeal to trigger migration to this service
offering. The magnitude of the migration is a matter of how significant the change is
compared to previous historical variations and customer preferences.
Within the model, it has also been assumed that there is no structural limitation to the
number of subscribers that can join one ISP. Although there is a physical break point
in the network access load, the extra flow of subscribers will be passed onto a
carrier’s network. The implication of a consequent migration flow from one, or more,
ISPs to another will appear in the Economic Module where interconnect fees heavily
weight the OPEX [6]. Additionally, a degradation effect is introduced into the QoS
ratio as a result of too many customers asking for support where the optimum service
has been under-dimensioned.
Delays within the model have been calibrated with current examples from the Internet
industry. Unlike in other industries, the response delay decreases as the marketplace
becomes more educated to the services and their functionality. Customer awareness
acts in the model as an attribute of delay in responding to the market stimuli. Once
again, the model relies on market research and analysis to ascertain which of the
reaction mechanisms within the model behave as in a real market situation.
Conclusion
The use of System Dynamics has proven to be valuable in mapping the dynamics of
the Internet into an interactive model. The model allows decision-makers to run
various scenarios to enable them to assess the implications of their tactics on long-
term network and cost issues. From an incumbent point of view, this simulation tool
can produce numerous and varied insights into the type of strategies that ISPs could
deploy as they face a wealth of tougher competition and cutting edge innovation. This
model provides a dynamic learning environment to understand the technological,
behavioural and strategic issues that we will increasingly have to face in the
information age.
Acknowledgements
I wish to thank Ann Matthews for her help and time as well as her understanding in
discussing this paper.
I wish to thank Mike Matthews for reviewing this paper and authorising its
publication.
Notes
[1] Best viewed in colour to aid clarity of the graphs.
[2] The Internet is to be understood as the digital platform for the global information
infrastructure that conveys both voice and data services.
[3] In contrast to market share, share of customer refers to the percentage of a
customer's/business' expenses that a company will be able to acquire during the
relationship established with them.
[4] In figure 5, the phenomenon has been amplified in its magnitude for clarity
purposes.
[5] In contrast to market share, share of customer refers to the percentage of a
customer's/business' expenses that a company will be able to acquire during the
relationship established with them.
[6] OPEX stands for OPerational EXpenditures as opposed to CAPEX, CApital
EXpenditure.
CC BY-NC-SA 4.0