1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Scenario Modelling of Demand for Future Telecommunications Services
Jeremy Barnes, Fraser Burton, Ian Hawker and Michael H Lyons
Systems Research Division
BT Laboratories
Martlesham Heath
Ipswich IPS 7RE, UK
Tel: +44 473 643210
Fax: +44 473 647410
E-mail: barnes_j_w_r@bt-web.bt.co.uk
Abstract:
It is widely believed that the world is entering the Information Age , and telecommunication
companies must make critical investment decisions based on how much information customers will
want to move in the future. Understanding and preparing for the range of possible customer demand
scenarios is vital for long-term success in an increasingly competitive market. However, detailed
forecasts are impossible to make since the market is as yet undefined. Scenario modelling is useful in
developing the understanding telcos.need to achieve this success.
We have developed a system model to i igate the effects of different business and
technological drivers on the demand for future telecommunications services, using the software tool
iThink. Drivers include the number of people teleworking and increasing computing power. These
interact to produce usage dynamics for generic services covering conversation, messaging and data
transfer, which are then used to calculate resulting network traffic.
Our results suggest that the key uncertainties are the rate of imp: in general IT sophisti
and the extent of teleworking. High growth in both of these produces rapid growth in peak traffic,
whilst low teleworking delays that growth. Slower improvement in IT sophistication severely limits
growth, since increasing computing power could stimulate large volumes of traffic. Small increases
in the use of video applications also produce significant traffic growth, and these factors combine to
give large uncertainties. The behaviour of this system is di d with to individual
business sectors, d ing system di ics as a useful h for i igati at
supply-demand systems.
Business Decision-Making, page 11
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Scenario Modelling of Demand for Future Telecommunications Services
1 Introduction
The tel ications market is ‘ing rapidly, with wid d ictions that demand will grow
by orders of magnitude over the coming decades (eg Gilder 1991, Lyons et al 1993a and 1993b), and
almost weekly Teports of new ions between npanies (telcos),
cable TV p isati and x ics giants. Telecoms operators
therefore face a complex environment experiencing rapid change, resulting in large uncertainties and
consequent risks associated with any course of action.
These elements suggest the demand for telecommunications services as an ideal problem for scenario
modelling using system d: We have developed a number of models aiming to explore
the risks iated with i ies for “broadband” (ie fast, high capacity)
networks and to analyse sensitivities to varying demand scenarios. Seminal work in system dynamics has
stressed the benefits of applying the techniques to investigate supply-demand systems (Senge 1990), and
the telecoms problem outlined presents an especially interesting example due to the long lead-times
(about 5 years) required for national roll-out of upgrades to the “access network” (the final link from the
network operator to its customers).
Here, we present an example of this work in which we have developed an iThink (Peterson 1992) model
to investigate demand scenarios for business customers in different industry sectors, based upon top-
level business and technological drivers. We have aimed to develop the modelling approach and
understanding of the system, rather than to formulate predictions, and our results show example
sensitivities.
We begin with a brief description of the background to this modelling work, and then introduce the
demand drivers investigated, with a description of their and i ios are then
developed to inty within the system, and top-level results from the model are
presented. We describe the application of these results to individual business sectors using simple
metrics to characterise the telecoms activity of a typical employee in each, and present sample results for
the usage of several generic services by people employed in different industries under alternative
scenarios.
We lude with a di ion of the und ding gained from the modelling process, identifying the
key uncertainties and | noting the broad scope for further applications of the drivers modelled. Topics of
ing work are d h the lication of results from the model to sample customer
data and visualisation of these results using a Geographical Information System (GIS). Finally, we
present ideas for further development of the modelling approach to improve ing of the whole
ipply-d d system for broadband networks and services.
2 Background
Toi igate the system d: ics app we decided to apply it to model the demand for broadband
telecommunications services in the UK market, with particular attention to the supply-demand system.
Initial work built from the classic analytical market diffusion models such as that proposed by Fisher and
Pry (1971) and Bass (1969). These models are based upon biological population studies (Mahajan and
Wind, 1986), assuming that the decision to purchase a product is an imitation process with a probability
proportional to the product’s market penetration. Bass divided the market to give 1-2% “innovators”
whose purchase probability is constant and “‘kick-starts” the diffusion process. We applied this approach
to telecoms services, realising that the degree of interconnection of a service is as important as its market
share (Lyons, 1994). An example of this is today’s videophone, where the full functionality is only
Business Decision-Making, page 12
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
available when calling another videoph a standard telephone can call a vi in the same way
as it calls another voice-phone, so it is fully interconnected.
A weakness of the Bass model is that it produces positively-skewed S-growth, as opposed to the
negatively-skewed curves found in real data (Simon and Sebastian, 1987). The positive-skew is due to
the continued influence of innovators through the diffusion process, whilst the negative-skew could be
due to a number of factors, including limited supply. These effects are difficult to treat analytically, but
are easy to simulate, and we investigated them taking the example of limited supply due to slow
deployment of a broadband network, and its effect upon demand for broadband services.
Supply and demand for network connections is only part of the story, though, and the next step was to
investigate demand for services and the i for network des. This led us to focus on
service demand drivers, which is the work we present here.
3 The model
One of our early objectives was to i igate the infl of external drivers on demand for telecoms
services by constructing scenario models (Wack, 1985 and de Geus, 1988), and, having investigated the
dynamics of the top level telecoms supply-demand system, iThink was again used to focus on this area.
Work to date has yielded some interesting results, and further development will incorporate these with
the effects of limited supply.
3.1 Top level dynamics & interactions
We have noted that significant growth is predicted for telecoms traffic. Fig.1 shows the top-level iThink
map of a number of drivers likely to influence this growth for business use, and a description follows for
each:
Fig.1 - Top Level Dynamics in the iThink model
{IT sophistication
paper into
Business Decision-Making, page 13
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
° ii logy IT isticati According to Moore’s Law (Greenop et al, 1993), the
ratio of | processing power to price doubles roughly every 18 months in the computing domain, but it is
uncertain whether we shall see an equivalent increase in the IT sophistication of the average person, at
work or at home,
°
Networking information (Electronic info): We assume that the greater the Proportion of
information that is available elect tonically, the greater the i ive to network
ie for publishers and other i P to move into the emerging electronic media market.
° Shift to teleworking: When a person traffic currently carried by npany private
exchanges (PBXs) and Local Area Networks (LANs) moves to public networks. This is modelled as a
modified Bass diffusion process.
° Substitution of tel for travel ti People already substitute telecoms for travel
whenever they make a phone call, and further substitution could occur as a result of telework and
displacement of business travel.
°
Substitution of video for plain voice: Further significant substitution of telecoms for travel
could require levels of interaction which can only be achieved using high-quality video links.
& of for . functional it Some proportion of current
are mainly ional and theref ible to substitution by i
of this are the growth of voice mail, and ordering a pizza by fax.
These demand dynamics are represented by fractions of the working population, except for IT
sophistication, which is measured in arbitrary units normalised to an initial value of 1. Most of them use
a diffusion equation, and the interactions between dynamics can be seen in Fig.1. Experience has shown
that in soft models such as this, it is easy to mi doubl t these infl so we
each chain of interactions as follows:° Technology - Improving technol will drive the migration
of information from paper and isolated el ic storage to ked media servers (modelled by “go
electronic”). Improving power:price ratios for tech could also stimul: I king and hence
a ings by providing terminal and network equipment.
° Ink
systems may be expected to facilitate telework, since
although techinology solutions are already available at realistic prices, difficulty of access to
information from remote locations remains a major limitation (Huws et al, 1990).
° Messaging - Information that is already networked is more likely to be communicated using
messaging, and flexible (remote) working could mean that people will share fewer common working
hours, making conversations less easy to establish and therefore encouraging people to use messaging for
functional communications.
° Video - If telecommunications are required to take the place of an increasing proportion of face-to-
face ings (eg, for 2)» imp: ig qualities of i ‘ion will be required. Continuing
advances in technological performance:price have recently given huge growth in the desktop multimedia
market, and both drivers are likely to contribute to an increa
sing p ion of video
These demand dynamics are featured in the iThink model, and then applied to individual business
sectors as described in §3.4.
Business Decision-Making, page 14
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
3.2 Scenarios
IT sophistication is the basic driver for the demand system. There is little doubt that technology will
continue to improve exponentially, but it is uncertain to what extent increasing computer power will
translate into increasing demand for public network capacity. This is the subject of much current debate,
for example in the question of whether “multimedia” will become a mass market in the near future (The
Economist, 1993).
Another topic of current debate is the likely future extent of teleworking. Some claim that the
combination of a growing amber of knowledge workers, plus economic and environmental
i d telework inevii . Others feel that the initial expense of home
equipment and network connections, plus practical difficulties i in changing working patterns may be too
much for most organisations to overcome.
The other dynamics are driven by IT istication and uptake of ‘king. If Moore’s Law doesn’t
apply for mass market IT sophistication, it is unlikely that there will be widespread teleworking in the
near future. But even with such sophistication, teleworking is by no means guaranteed to become the
norm - other factors may well prevent it.
From this di ion, we can the
1 “High Growth”: IT sophistication follows Moore’ 's Law, and high latent demand for teleworking
2 “Medium Growth”: Moore’s Law growth in IT sophistication, but low demand for teleworking
3 “Low Growth”: Slow improvement in IT sophistication and (resultant) low teleworking
Fig.2 - Top Level Dynamics
a- High Growth (IT sophistication doubles every 18 months, ie Moore’s Law growth, and high
teleworking)
100 2
90
—— : relative IT
80 sophistication
A [nar 2: % electronic
60 info
=
ZS sot Fo php bY natn 3: % teleworkers
>
40 4: %
30 aera 5: % video comms
20 6: % messaging
25
Business Decision-Making, page 15
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
b- Medium Growth (IT sophistication doubles c - Low Growth UT sophistication doubles every
every 2.5 years, but low teleworking) 5 years, but high teleworking)
100 100 pone
t 2
90 t
| 0 /
80 8 i
70 ze /
60 . 8 /
2 =
= so = 50 ‘
5 7 40 / —s
é a a
30 a0 / 6
/
2 20 y ;
10
ao ° == sues
8 oe 2 2 & 8
Yoars
3.3 Top level results
The top level results for the three scenarios are shown in Fig.2. A long (25 year) timescale is shown
deliberately i in these graphs to illustrate longer-t trends. It is imp to that the obj
is to igate demand ios, not to make predi
The effect of the different dynamic interactions modelled can be seen clearly in these results. In all cases,
the initial driver is electronic information, giving an increase in messaging. Increasing telework and
hence telemeetings and video comms also result in the high and low growth scenarios (high latent
demand for teleworking), although in the low growth scenario, the growth in electronic information is
much slower due to low growth in IT sophistication.
In the high and medium growth ios, improving IT ‘istication also has direct effects as its
growth rate accelerates. This is first seen in increasing ‘telework, which begins to grow rapidly from year
15 even in the medium growth (low telework) scenario, entirely due to improving technology. Under
high growth, telework accelerates from year 8 after being jump-started by high latent demand.
Obviously, telemeetings and hence video communications follow the telework dynamic rather closely,
and telemeetings, telework and messaging approach saturation at year 25 as we have assumed some
practical limits to growth in these applications.
After the growth in the 3-7 year period already discussed for high and medium scenarios, messaging
plateaus, and further growth is also dependent upon increased telework. This takes effect from about
year 12 under high growth, and from year 17 under mediums
The low growth scenario shows no effect of IT sophistication within the 25 year period, other than as a
of i ic info (which is slower than under the other scenarios). This scenario
is identical to high growth, but with the first 7 years stretched to fill 25.
3.4 Application to individual business sectors
To be applied to sample customer data, the top-level dynamics must be related to service usage by
customers in different business sectors. A small number of generic services were chosen for modelling,
and the influence of the top level drivers upon the usage of each service within different business sectors
Business Decision-Making, page 16
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
has been assigned using four metrics to characterise each sector. The peak traffic generated by a
customer in each business sector under the different scenarios outlined can then be calculated. Business
sectors are identified using the first digit of the UK government’s 1980 definition of Standard Industrial
Classification (SIC), giving the 10 sectors shown in Fig.3 - Agriculture (SIC 0) through to Government
(SIC 9).
The model aims to relate the top-level technological and market drivers to the usage of five generic
services:
° voice telephony
° video telephony
° fax
° email
° — data transfer
We have led the service d ics identi for each business sector, with the influence of each
top level dynamic depending on four simple metrics used to characterise each SIC. The variation of these
metrics by business sector is shown in Fig.3:
for more than just word-processing
° — IT intensity - the p of empl who use a
(internal source). .
Information intensity - the p of emp who are in information-based
(Employment Dept LFS, 1993)
° — Interaction intensity - the p of emp! who are in i ion (eg meetings) i
ions (eg sales ives, some clerical... - Employment Dept LFS, 1993)
Travel intensity - the average number of journey stages per employee per year (a “journey
stage” is a standard travel metric used by the Department of Transport to measure journeys made
for the purpose of transporting people - Transport Dept, 1991)
Fig.3 - Metrics to characterise business sectors
% IT users Co % info % meatings
workers people stages
90 3.0
=
80 os 3
70 BZ ° z
Z
60 Z PAr 2:0 Hy =
3 s0 Z Z| az
4] Z Zl e
£ is Z| Z A 1.5 gs
wo: Z Z, Z =
30 Z Z Ar 1.0 ge
Z Z Z 2
Z Z Z =
10 Y Z Z| H
° Z & Zl o.0
Business Decision-Making, page 17
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
To illustrate the approach taken in applying the top level dynamics to service usage within business
sectors, the equations modelling the use of conversational services (ie voice + video telephony) are as
follow:
teonversion = fmeeting_people F relemeetings(fieteworkers Fav meeting + Mav,journeys tav.remote_meeting)
where frieworters = 1/3(Fime_ouof.opice + Finfo_workers + frrworters )Freteworkers aNd
Fctemeetings is the top-level fraction of telemeetings
Freeteworkers is the top-level fraction of teleworkers
Singo_workers is the information intensity
Srr_worters is the IT intensity
Smecring_people is the interaction intensity
Nav journeys is the travel intensity
Tav.meetings is the average manager’s 3-4 hours per day spent in meetings (Young, 1986)
Tav:remote_mectings is the average length of a meeting away from base (estimated at 2 hours)
Sime _out_of. office is the fraction of time spent away from the office (from travel intensity)
Thus, a Proportion of the tiie currently spent in face-to-face meetings at and away from the normal
workplace is d to 1g to the top-level dynamics for telework and
telemeetings, modified by the interaction intensity, travel intensity and fraction of teleworkers in a
business sector. The fraction of teleworkers is determined by the top-level telework dynamic, adjusted by
the information, IT and travel intensity for a business sector. Some of this total conversation is then
to ding to the top-level for and ion, modified
by the ratio of information:interaction intensity for a business sector. The remaining conversation time is
then split between voice and video according to the top-level dynamic for video, adjusted by IT intensity
for the sector.
A similar approach is taken for messaging (fax + email) and data transfer, and bit rates for each service
type are applied to calculate the total traffic per person employed. Busy hour fractions for each service
are also used to calculate peak traffic per person, and some sample results are given in the following
section.
3.5 Results for service usage by business sector
Fig.4 shows the relative growth of service usage in the Construction sector (SIC 5) and in Banking and
Finance (SIC 8) under the high growth scenario over 25 years. We emphasise that these traffic figures
are estimates based upon sample traffic data, rather than hard figures for current network usage, but
plotting traffic rather than direct service use immediately shows the effect of even a rather small amount
of video (hence the logarithmic scale). This is because a second of video at 2Mb/s gives about 30 times
the traffic of a second of voice (64kb/s).
Referring again to Fig.3, it is apparent that Construction and Banking & Finance are interesting sectors
to compare, since SIC 5 has the lowest proportion of IT users, information workers and meetings people,
but the highest travel intensity. In contrast, SIC 8 has the highest IT users and information workers, and
amongst the highest meetings people, with slightly above average travel intensity. The effects of these
differences can be seen from the results in Fig.4, most obviously in the difference between the two
Business Decision-Making, page 18
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
sectors in orders of magnitude for each type of traffic, but also in the dynamics themselves, which we
now discuss.
The most obvious contrast in the results is perhaps in the difference in video traffic, which exceeds
voice traffic by year 5 for SIC 8, but not until year 17 for SIC 5, whilst in both cases voice grows at a
similar rate. This is a matter of scaling, and is entirely due to the difference in IT intensity between the
two sectors. The shape of the video curve in each case is identical (cf the video curve in Fig.2a), since
the IT intensity (in common with the other three metrics) is a constant, and this is raised as a subject for
further investigation. Note again that a small increase in video usage produces a large increase in traffic,
and that by year 25 in SIC 8, the average time spent using video is still only 60% of that for voice. In
both sectors, voice has saturated by year 25, and in SIC 8 is actually declining from year 22 onwards.
Rather more ii ing are the ‘ics for ing (fax and email), which we have seen are
strongly dependent upon both the IT intensity and the extent of teleworking within a sector, which is
itself d dent upon the i ion and travel intensity. The immediate observation is that
email substitutes fax rather more quickly i in SIC 8, with fax vanishing by year 13 compared with year 16
in SIC 5. N ile, total traffic i to grow in both sectors throughout the 25 year
period, and it is worth noting that although email is a much more efficient medium than fax for text
transmission (requiring some 16kb per: uncompressed A4 page compared with 300kb for a fax), email
systems i ly support ing. This is likely to lead to significant increases in
email traffic, which is modelled by relating the email bit rate to the penetration of video at the top level.
Of particular interest are the differing shapes of the email curves, with the distinct knee at year 7 for SIC
5. A similar effect can just be detected for SIC 8 a little earlier, and this is again due to the large
difference in IT intensity between the two sectors, resulting in a greater delay in transferring to email in
the Construction sector. Deeper investigation also shows that telework obviously grows much more
quickly and reaches a higher penetration in SIC 8, driving this effect. Comparing these email curves with
the top-level messaging dynamic in Fig.2a suggests that much of the growth in messaging is in new
email traffic, and the final substitution of email for fax (which occurs over just 2 years in both cases) is
combined with the second phase of messaging growth from years 12-17 to give especially strong growth
in email over the second half of the 25 year period.
Growth in data is driven directly by top-level IT sophistication, with a delay ing on the IT
intensity of the business sector, so both sectors show the expected exponential growth. Data is very
much an uncertainty at the top level, and could be the only significant factor in extreme scenarios.
Fig.4 - Traffic growth per person for contrasting business sectors under the High Growth scenario
a- SIC 5: Construction
= 100000
*% 10000
& 1000
5 100
s 10
Fa
2 1
ZB 04
2 0.01
= 0.001
°
Business Decision-Making, page 19
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
b- SIC 8: Banking and Finance
z= 100000
> 40000
&
c
a
€
5
a
&
2
=
4 Di jion & Future D
We have shown how we have applied system dynamics techniques to model demand for telecoms
services from different business sectors. The modelling approach has been useful in structuring the
problem in terms of a small number of top-level drivers and some simple metrics characterising
individuals in different business sectors. We shall now discuss the learning gained from the modelling
work, based upon the results presented in §3.3 (Fig.2), §3.5 (Fig.4) and in Fig.5, which shows the
average traffic growth across all business sectors under the three scenarios proposed.
Fig.5 summarises the task faced in i igating future tel demand. In planning future
networks, telcos need to know how traffic is likely to grow, but the range of uncertainty seen in the
results from these simple scenarios illustrates that this task can be extremely difficult, especially over the
required for ive network :
Fig.5 - Peak traffic per person employed averaged across all business sectors
0.84 High
Peak traffic per person (Mb/s)
System |
offers an ive approach. Having ii d the range of possible future
and their likely usage, the problem has been to investigate the factors likely to
affect the rate of growth towards this possible future level of demand. We have achieved this by
Business Decision-Making, page 20
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
modelling a small number of drivers, and by constructing simple scenarios based upon varying two of
these drivers.
In developing the scenarios in §3.2-3.3, we noted that the key uncertainties seem to be the rate of growth
in general IT sophistication, and the rate of growth in telework. The effects can be clearly seen in Fig.5,
where high growth is driven by Moore’s Law and early demand for telework, medium growth is the
same, but delayed by low initial teleworking, and low growth is slow due to IT sophistication advancing
far less than under Moore’s Law roughly following the current trend for telephony traffic, doubling
every 4 years or so (O’Mahoney, 1994). The rapid growth rates obviously mean that the scenarios are
sensitive to any delays in the system. In §3.4-3.5, we showed that the IT intensity of a business sector
could be a key factor in determining future growth in telecoms demand, and we noted that any growth in
video usage is particularly significant due to the vastly higher data rates required. These two points are
obviously important in identifying customers likely to require early network upgrades, and also in
ping for ing new growth.
The flexibility of the modelling approach has also been seen as valuable. Placing the dynamics at the top
level means that the understanding gained can also be applied to other problems, for example in
investigating demand from residential customers. In future work, we plan to further develop our
understanding at this level by investigating drivers including costs for travel and accommodation,
changing employment patterns and average leisure time. However, the decision to fix as constants the
metrics used to characterise business sectors, although simplifying the model, means that the effects of
changing working practices in particular sectors were not investigated as part of this work. Whilst it is
possible that, for example, the relative IT intensity of different sectors may remain constant, so that the
top-level dynamics can reasonably be applied across all sectors, this is a special case, and it would be
useful to be able to investigate the effect of intervening to stimulate such a change in a particular area of
business.
Other topics of continuing work include investigating the effect upon demand of alternative broadband
and the application of results from the model to sample customer data, using a
Geographical Information System (GIS) for visualisation. Initial results from using the GIS and other
visualisation tools have been encouraging, but could increase the risk of output from scenarios being
misunderstood as firm predictions (despite cautions from the modelling team; New Scientist, 1994). We
therefore believe that the of system d: ions will require ongoing effort to
enable decision makers to embrace uncertainty as a constant companion demanding respect, and to
support the structuring and investigation of problems in ways which foster learning and reasoned
responses.
5 Conclusion
We have described the application of system i to i i the supply-d id
system for telecommunications services, using the software tool iThink. System | dynamics has been seen
as a useful approach to this problem, given the complexity and rate of change facing established telcos,
resulting in large uncertainties and consequent risk.
Our emphasis has been upon investigating the modelling approach for this system and building
understanding, rather than making predictions. We have described the development of simple scenarios
based upon top-level dynamics, and results from these scenarios have been applied to the use of five
generic services in different business sectors. Sectors were characterised by simple metrics, and the
traffic for each person employed has been used to calculate the growth in peak traffic under the
alternative scenarios.
In discussing this growth, we have noted the significance of even a small increase in video use, which is
driven by improving IT sophistication in the model. This means that the results are sensitive to delays in
Business Decision-Making, page 21
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
this improvement, and indicate it as a key uncertainty. The extent of teleworking is also suggested to
strongly influence these delays, and the relative IT intensities of different business sectors are noted as
another possible point of leverage. The broad application of the approach has been ised, and we
have identified possible d: ics for further i igation at the top-level. Other future work may use
isualisation systems to aid interpretation of results from these models, but we have emphasised that
such applications of system dynamics must continue to facilitate scenario analysis rather than claim
Delphic knowledge of future events.
6 References
Bass, F.M. 1969. New Product Growth Model for Consumer Durables. Man. Sci 15 215
The Economist . 1993. Multimedia’s Yellow Brick Road. 4 December.
Eaployment Dent Labour Force Survey. 1993. Standard Industrial Cl ion by Standard
i ion for all employed people in UK, May 1993 (supplied by Quantime Ltd)
Ha Cu petits Pry. 1971. A Simple itution Model of Te logical Change. Technol.
Forecasting & Social Change 375
de Geus, A.P. 1988. Planning as Learning. Harvard Business Review. March- -April 1988: 70
Gilder, G. 1991. Into the Telecosm. Harvard Business Review March-April 1991: 150-161.
Greenop, D., l-Pearson and T.Johnson. 1993. Broadband - Liberating the Customer. BTEngineering 12 4:
255
Huws, U., W.Korte and S.Robinson. 1990: Telework: Towards the Elusive Office. Wiley
Lyons, M.H. 1994. Modelling the Interactions Between New Telecommunication Sertvices. BT
Technology Journal. 12 (2): 109-113.
Lyons, M.H., M.Gell and P.Cochrane. 1993a. Telecommunications in the Third Millenium. CBTA Conf.
Proc., TeleCon ‘93: 141-146.
Lyons, M.H., K.O.Jensen and I.Hawker. 1993b. Traffic Scenarios for the 21st Century. BT Technology
Journal 11 (4)
Mahajan, V. and Y. Wind (eds). 1986. Innovation Diffusion Models of New Product Acceptance.
Cambridge, MA: Ballinger
New Scientist. 1994. Lies, damned lies and slick graphics. 5 March: 11.
O’Mahoney, M. 1994. MSc in Telecoms Engineering (course notes). University of London.
Peterson, S. 1992. Software for Model Building and Simulation. European Journal of Operational
Research 59 (1): 197-202.
Senge, P.M. 1990. The Fifth Discipline. Doubleday.
Simon, H. and K.H. Sebastian. 1987. Diffusion and Advertising: The German Telephone Campaign.
Management Science 33 : 451
Transport Dept. 1991. National Travel Survey Data on Business Mileage . Bespoke query.
Wack, P. 1985. Scenarios: Uncharted Waters Ahead. Harvard Business Review. Sept-Oct 1985: 72
Young, A. 1986. The Manager's Handbook . Sphere.
Business Decision-Making, page 22
