Analysys - 3G Mobile LRIC Model Doc for Sweden

>

Paper for National Post and

Telecom Agency (PTS)

Mobile LRIC model conceptual design and

model specification for 2G and 3G mobile networks in

Sweden

25 January 2008 303‐46

Analysys Consulting Limited

St Giles Court, 24 Castle Street

Cambridge, CB3 0AJ, UK

Tel: +44 (0)1223 460600

Fax: +44 (0)1223 460866

[email protected]

www.analysys.com

303-46

Contents

1 Introduction 1

2 Conceptual approach 2 2.1 Overview of conceptual issues regarding a mobile LRIC model 2 2.2 Selection of conceptual issues that require additional consideration 3

3 Conceptual issues relating to operator 5

4 Conceptual issues relating to technology 6 4.1 Spectrum allocations 6 4.2 Radio technology standards 7 4.3 Infrastructure sharing 12

5 Conceptual issues relating to services 15

6 Conceptual issues relating to implementation 17 6.1 WACC 17 6.2 Depreciation method 17 6.3 Mark-up mechanism and network externalities 18 6.4 Increments 18 6.5 Years to consider 20

7 Model specification 22 7.1 Network loading 23 7.2 Coverage parameters 24 7.3 Equipment capacities 26 7.4 Demand inputs and projections 27 7.5 Network topology 30

Annex A: Glossary of technical abbreviations Annex B: Revisions to approach following industry consultation

Copyright © 2008 Analysys Consulting Ltd has produced the information contained herein for the National Post and Telecom Agency (PTS). The ownership, use and disclosure of this information are subject to the Commercial Terms contained in the contract between Analysys Consulting Ltd and PTS.

Mobile LRIC model conceptual design and model specification for 2G and 3G mobile networks in Sweden | 1

303-46

1 Introduction

Analysys Consulting Ltd. (‘Analysys’) has been commissioned by the National Post and Telecom Agency (PTS) to perform an upgrade of the LRIC calculation model originally developed in 2003 by Analysys and HiQ (the ‘original model’) in order to take into account recent market developments. In particular, the upgrade will incorporate the emergence of 3G networks and services.

Since the development of the original model, the volume of mobile network traffic in Sweden has increased and 3G networks have emerged. In order to assess the potential requirement for further regulation of mobile termination, PTS needs a detailed understanding of the costs associated with the current market situation and the status of the networks in Sweden. This is to be accomplished by carrying out a set of upgrades to the existing 2G LRIC calculation model to incorporate the evolution in the mobile market, and 3G deployments and services.

This paper sets out two aspects of the upgrade process that need to be considered: its conceptual approach, and its proposed model specification. Certain elements of these two aspects are determined by the original model, and PTS does not intend to reopen the discussion on these matters. However, other elements will require fresh consideration – either as modifications to the existing 2G LRIC model approach, or as completely new issues. Following submissions from the Swedish industry parties on a consultation version of this paper (dated 26 October 2007), a number of modifications to the proposed approach have been made. These modifications have been included at the relevant points throughout the present document, and are collated in Annex B.

The conceptual design (Conceptual Design Final Version.pdf) and model specification (Model Specification Final Version.pdf) applicable to the original model can be found on PTS’s website.1

The remainder of this document is structured as follows:

• Section 2 describes the conceptual issues associated with the mobile LRIC model and distinguishes between those that are determined by the original model and those that require additional consideration

• Section 3 describes the conceptual issues relating to the operator modelled • Section 4 describes the conceptual issues relating to the technologies used • Section 5 describes the conceptual issues relating to the services provided • Section 6 describes the conceptual issues relating to the implementation of the upgraded model • Section 7 describes the proposed specification of the upgraded model • Annex A contains a glossary of the technical abbreviations used in this paper • Annex B lists the revisions made to the approach following industry consultation.

1 http://www.pts.se/Archive/Documents/SE/Conceptual%20Design%20final%20version.pdf

http://www.pts.se/Archive/Documents/SE/Model%20Specification%20-%20Final%20version.pdf


303-46

2 Conceptual approach

This section first provides a brief overview of the conceptual issues associated with the design of a mobile LRIC model. These issues were considered carefully during the conceptual design of the original model in 2003, and the decisions that were taken at that time are summarised briefly. The section then distinguishes between those issues which, in the view of Analysys, do not need to be considered again, and those which do require fresh examination.

2.1 Overview of conceptual issues regarding a mobile LRIC model

On 5 June 2003, PTS issued the final version of the Mobile LRIC Model Conceptual Design, which specified the basis on which the original model was implemented and applied. That document classified the conceptual issues in terms of four dimensions: operator, technology, service and implementation, as summarised in Figure 2.1 and described in more detail below.

Modelling conceptual issuesTechnology Service

Operator

Implementation

Figure 2.1: Framework for classifying conceptual issues [Source: Analysys]

Operator The characteristics of the operator that is used as the basis for the model represent a significant conceptual decision, with clear costing implications. The following choices are significant:

• What structural implementation of the model should be applied (top-down, bottom-up, hybrid)?

• What type of operator should be modelled – the actual operators, an average operator, or a hypothetical new entrant to the market?

• How efficient an operator should be modelled – individual operators, the most efficient in Sweden, the most efficient in the world, a hypothetical most efficient operator?

• What size of operator(s) should be modelled – actual size, average size, or some other size of hypothetical operator?


303-46

Technology The nature of the network to be modelled depends on the following conceptual choices:

• What radio technology standards should be deployed? What are the effects and existence of migration and other proxies for technology evolution?

• What is the nature of the spectrum allocated to the modelled operator (amount, band and fees paid)?

Service Within the service dimension, there are questions regarding the scope of the services being examined:

• What service set does the modelled operator support? • Are costs calculated at the wholesale or retail level (i.e. is the network

standalone or vertically integrated)?

Implementation A number of implementation issues must also be resolved to produce a final cost result. They are:

• What is the weighted average cost of capital (WACC) for the modelled operator?

• What depreciation method should be applied to annual expenditures (HCA, CCA, tilted annuities, economic)?

• What increments should be costed, and should they be marginal, incremental or average?

• What year(s) should results be calculated for? • What mark-up mechanism should be applied to costs that are

common to the increments? Should Ramsey pricing be employed? • Is a network externality surcharge appropriate?

2.2 Selection of conceptual issues that require additional consideration

In Table 2.1 below, the issues outlined in Section 2.1 are listed, together with the decisions made for the original model. There are a number of issues that Analysys does not consider require revisiting for the purposes of the model upgrade: the decisions regarding these issues that were taken by PTS in the original conceptual design are considered independent of the refinements being proposed for the model upgrade. The remaining issues in Table 2.1 have been marked with “yes”, indicating that Analysys believes that they that require further consideration in the context of upgrading the mobile LRIC model. Note that the derivation of the WACC for the upgraded model is an issue that may be studied by PTS in a separate analysis and therefore does not fall within the scope of the present paper.

The following sections of this Paper discuss these issues under the headings of operator, technology, service and implementation, and specify the proposed approach.


303-46

Conceptual issue Determination for original model Reconsider?

Structural implementation of model Hybrid (bottom-up reconciled against top-down) No Type of operator Actual operators No Efficiency of operator All operators are efficient by Swedish standards No Size of operator Actual size No Radio technology standards GSM only Yes Spectrum allocations Actual No Service set Voice and data No Standalone network or vertically integrated Full business costs No WACC Use Capital Asset Pricing Model (CAPM) No

Depreciation method Economic depreciation No Increments Two; traffic and subscribers Yes Type of increments Average No Years to calculate results for All past and future years Yes Mark-up mechanism Equi-proportionate mark-up (EPMU) No Inclusion of network externality No No

Table 2.1: Decisions on the conceptual issues taken for the original model, and items requiring

further examination [Source: Analysys]


303-46

3 Conceptual issues relating to operator

This section discusses the following conceptual issues:

Conceptual issue Determination for original model Reconsider?

Structural implementation of model Hybrid (bottom-up reconciled against top-down) No Type of operator Actual operators No Efficiency of operator All operators are efficient by Swedish standards No Size of operator Actual size No

Table 3.1: Operator conceptual issues in the original model [Source: Analysys]

The original model was a hybrid model of the three actual mobile network operators in Sweden at the time (TeliaSonera, Telenor and Tele2). It considered them all to be efficient by Swedish standards, and costs were assessed on the basis of each operator’s actual size, plus a forecast of converging operator size. In the original model, the network operators carried a certain volume of traffic for independent service providers and MVNOs, and this volume of traffic was considered intrinsic to the host operator’s own volume of traffic (and therefore its costs). In the upgraded model, service provider and/or MVNO volumes will still be considered within the host operator’s volume.

Recommendation 1: Maintain the established operator principles for the basis of the cost calculation: a hybrid model of the actual operators considering each at its own actual size including specific traffic types they each carry through the support of MVNOs (with converging forecast) and applying the standard that each operator should attain reasonable Swedish standards of efficiency.


303-46

4 Conceptual issues relating to technology

This section discusses the following conceptual issues:

Concept Determination Reconsider?

Spectrum allocations Actual No Radio technology standards GSM only Yes

Table 4.1: Technology conceptual issues in the original model [Source: Analysys]

This section first considers the conceptual determination regarding spectrum allocations, and propose that it should be left unchanged. It then discusses which radio technology standards should be used in the updated model, and considers the issues arising from infrastructure sharing.

4.1 Spectrum allocations

The original model calculated mobile network deployments using the actual amount and frequency of GSM spectrum that each operator had. The model also reflected the annual GSM licence fees paid by each operator to PTS. The reason for this was that the use of 900MHz and 1800MHz spectrum for 2G services is different for each operator in Sweden. An operator’s spectrum allocations may lead to different network deployments and therefore to different levels of costs.

The W–CDMA (UMTS) licences issued in 2000 for 2100MHz frequencies were awarded in a beauty contest, rather than an auction. Each of the licences was for 2×15MHz + 5MHz of spectrum in the same frequency band, with fees comprising an annual revenue charge and transmitter fees.

Given that the UMTS licence allocations are currently identical in size and in the same frequency band (2100MHz) it is appropriate, for consistency with the original model, to implement actual UMTS spectrum allocations within the upgraded model.

A further potential aspect that may arise in future years is that of spectrum liberalisation, whereby spectrum traditionally used for GSM technology can be re-used for UMTS technology instead. However, since the current objective of the model is to consider actual spectrum usage, and GSM spectrum is currently tied up in the existing networks, this would only be implemented once it was intrinsic to the operators’ actual cost base.

Recommendation 2: Model actual spectrum allocations, in this case for GSM and W–CDMA (UMTS) technologies at 900MHz, 1800MHz and 2100MHz accordingly. Given the long-term uncertainty in the availability and use of 900MHz for each mobile operator, PTS intends to explore the impact of this issue on network costs: whether existing operators re-acquire their existing spectrum allocations, whether existing 900MHz operators lose some


303-46

or all of their spectrum in this band, or whether an operator previously without 900MHz spectrum receives a share of such frequencies. Spectrum fees for radio transmission (e.g. per transmitter) and for microwave links should also be included in those network element costs.

4.2 Radio technology standards

The original model only considered three operators and their GSM operations, although GPRS was included as an enhanced second-generation technology. Analogue (NMT) technology was not included. More significantly, UMTS technology was not included: although the UMTS licences had been awarded by this time, the deployment and evolution of UMTS networks were seen to be too uncertain to be included. However, in the present day, all three major existing GSM operators have deployed UMTS networks in some form, and another player has entered the market offering UMTS services. In particular:

• Hi3G launched the first Swedish UMTS operations in 2003 • Telenor took over Vodafone’s UMTS licence as part of the acquisition in 2005 and has

continued with UMTS deployments: it has in fact entered into an infrastructure-sharing venture with Hi3G

• Tele2 and TeliaSonera have launched a UMTS network under a joint venture called Svenska UMTS AB.

The subscriber base for Hi3G in particular is now significant, standing at over 400 000: this was not captured in the original model. Also, UMTS network traffic is rapidly increasing in volume, as shown in Figure 4.1 below, with 12.7% of voice traffic now originating from UMTS subscribers. In addition, packet-switched data services (including HSPA) are important emerging services carried on the UMTS network.


303-46

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1H 2004 2H 2004 1H 2005 2H 2005 1H 2006 2H 2006

Out

goin

g m

inut

es fr

om U

MTS

(billi

ons)

Figure 4.1:

UMTS-originated

outgoing traffic on a half-

year basis, 2004–06

[Source: PTS]

Since migration to UMTS is now a significant factor in the market, the original model diverges from reality in two important respects:

• the cost base of TeliaSonera, Telenor and Tele2 is now a weighted-average blend of GSM and UMTS networks, with UMTS set to take an increasing share in future years

• Hi3G now carries material volumes of mobile-terminated traffic using a UMTS-only network (supplemented by a national roaming agreement with Telenor).

Recommendation 3: W–CDMA (UMTS) technology should be incorporated within the upgraded model and the actual costs for combined GSM+UMTS operators should be based on the multi-technology situation that they are currently facing.

Since late 2006, all operators have launched HSPA services in parts of their UMTS network. HSPA technology has the following characteristics:

• it is a data-only upgrade to UMTS which allows the same radio spectrum to be used for improved packet data services (e.g. by using a different modulation and channel coding scheme, which requires support in the handset receiver)

• it utilises the same equipment (channel elements and NodeB infrastructure) used for UMTS voice traffic

• the volumes of data traffic that may eventually be carried on HSPA bearers is unknown.

Swedish UMTS networks will certainly be used to carry voice services in the long term, whilst messaging, video calling and Release-99 (‘normal’) packet data services represent additional non-voice services that the networks support. High-speed data technology – HSPA – is currently being rolled out across the network, and services are available. The volume of non-voice traffic carried by UMTS in the long term is, however, uncertain.


303-46

Recommendation 4: The model should reflect Release-99 voice and data services, as well as the inclusion of HSPA technology and services, in order to explore the impact of both normal and high-speed non-voice services (and their volumes) on the cost of voice traffic. The model will investigate the degree of variation in voice costs which may arise because of the uncertainties in non-voice (particularly HSPA) traffic.

As a result of including W–CDMA technology, the model will need to take account of the fact that there is now migration from GSM to UMTS for subscribers, voice traffic, messaging and data traffic. Swedish mobile networks are, in reality, going through a complex multi-technology cycle, in which any particular generation of mobile network technology is only expected to remain valid for a finite number of years. However, the original model, or any long-run cost model, is effectively making predictions of parameters in perpetuity. Therefore, as mobile operators manage the migration of demand and subscribers from one generation of technology to the next, so should the model make corresponding adjustments to its parameters. Therefore a multi-technology approach should be adopted for the upgraded model.

These effects apply to both GSM+UMTS operators and to the UMTS-only operator. Both can, it is suggested, be characterised as being within a multi-technology cycle, although in the case of the latter, the future next-generation technology it is moving towards is as yet unspecified. GSM networks are likely to achieve up to, or in excess of, 20 years of operation (for example 1993–2013) and the same can be expected of UMTS (for example 2003–2023). From this it is clear that the UMTS-only operator is only a few years into its estimated 20-year technology lifecycle, whereas the GSM operators are approaching the mature stages of the cycle. In long-run modelling terms, this means that the long-run costs of a UMTS-only operator is more dependent on (potentially more uncertain) forward-looking forecasts. However, the principles of technology generations should, it is proposed, be treated consistently for both GSM and UMTS lifecycles.

Recommendation 5: Model GSM and UMTS technologies taking into account the real-world effects of operating in a multi-technology environment. These effects include: migration of traffic to the next generation of technology; changes in demand and/or network cost levels between generations; and the sharing of certain network and overhead cost elements across technology generations. The mobile operators are subject to a degree of real-world uncertainty regarding the long-term outcome for GSM and UMTS networks. As such, the model will also investigate the cost implications of a number of possible situations: (a) GSM persisting in the long term to support background telemetry services, some mobile subscribers remaining on GSM, and fall-back voice and data national roaming for UMTS handsets; or (b) GSM being fully replaced by UMTS, possibly with the inclusion of UMTS at 900MHz.

The three areas that are significant in the present context of the costing of mobile termination are migration of traffic, generation changes, and economies of scope. These are discussed below.


303-46

4.2.1 Migration of traffic

The migration of traffic from one network to another affects the profile of the output produced by the network assets of each technology generation. This changes the level of unit costs over time for each generation, irrespective of the chosen depreciation method.2 The long-run cost from a single technology that can be operated in perpetuity will be lower than the long-run cost of a technology with a finite lifetime. This is because, for a technology with finite lifetime, traffic levels on the network will reduce3 in the later years and some assets (e.g. licence fee, common costs) will not be perfectly down-scalable with reducing traffic volumes. However, a single-technology model will not necessarily capture any productivity gains from moving to the next technology, such as higher system capacity or a corresponding greater service demand. Therefore, a single-technology, long-run cost may be higher than the blended average cost from a sequence of multiple generations of a mobile cellular technology.

From a cost modelling perspective, it is important to understand the implications of modelling the demand of network technologies – and their generations – on the level and timing of cost recovery. Figure 4.2 provides a schematic illustration of the potential output profiles.

Time

Time

Out

put

Out

put

Single generation in perpetuity

Multiple generations

Figure 4.2:

Possible output profiles

for single-generation and

multi-generation models

[Source: Analysys]

2 Although, of course, the choice of depreciation method determines when and how unit costs change as a result of migration.

3 Planned or unplanned reductions in traffic levels.


303-46

4.2.2 Generation changes

In order to capture the effects of successive technology generations, factors which change from one generation to the next may be defined in the model. As discussed in the previous subsection, successive generations of cellular technology can be expected to have increased output. Also, the cost per unit of capacity is likely to reflect continued technological improvements in the manufacturing of network equipment. The nature of these evolutions of demand and cost is illustrated schematically in Figure 4.3 below. The key issue for a LRIC model is consistency: it would be inappropriate to model continual levels of demand growth without technological evolution (and vice versa).

Uni

t cos

t of c

apac

ity

Time

GSM

UMTS

Serv

ice

dem

and

Time

GSM

UMTS

Figure 4.3:

Demand and cost proxies

for multiple generations

[Source: Analysys]

Since the GSM network may be scaled down (and perhaps even shut off) during the course of the modelling period, a retirement algorithm will need to be included in order the model. For consistency, it would be appropriate to include this for the UMTS technologies as well.

4.2.3 Economies of scope

A number of network and non-network costs will effectively be shared by successive generations of technologies – in these instances it will be possible to extract the same (or greater) utilisation


303-46

from an asset irrespective of the rate or existence of migration. Certain network assets fall into this category: for example, base station sites may continue to be rented from one generation to the next, backhaul transmission may be transparent to GSM and UMTS traffic, business overhead functions may support both technology generations, and so on. Given these economies of scope between technology generations, service costing for certain assets should be independent of migration.

4.3 Infrastructure sharing

The original model did not need to consider infrastructure sharing, since there was no significant sharing of GSM infrastructure in Sweden. However, once UMTS is included in the model, this becomes an issue because, although the four main network operators are independent, there is now a high degree of infrastructure sharing between them. This requires a more complex network costing approach. In addition to infrastructure sharing between an individual operator’s own GSM and UMTS networks, there are three instances of infrastructure sharing between operators that need to be taken into account:

• Hi3G’s national roaming agreement with Telenor, which was originally signed with Vodafone in 2002

• UMTS infrastructure sharing between Hi3G and Telenor in a joint venture called 3G Infrastructure Services AB

• UMTS infrastructure sharing between TeliaSonera and Tele2 in the joint venture Svenska UMTS AB.

The situation regarding infrastructure sharing between the national networks in Sweden is illustrated in Figure 4.4 below.


303-46

GSM

Own UMTS

Own UMTS

Shared UMTS

GSM

Shared UMTS

GSM

Geographic coverage

TeliaSonera

Svenska UMTS

Tele2

Telenor

Hi3G

3G Infrastructure Services

Geographic coverage

Infrastructure sharing between TeliaSonera and Tele2

Infrastructure sharing between Telenor and Hi3G

Independent operator Joint venture

KEY

GSM national roaming agreement with Telenor

Figure 4.4: Infrastructure sharing between Swedish mobile operators [Source: Analysys]

In order to model these instances of infrastructure sharing, it is proposed that the same stepwise approach be followed in each case:

1. Calculate the economic costs of each shared network on the basis that the total expenditures incurred for that network support the combined traffic volumes of the two parties. This gives the economic shared network unit costs of traffic, which can be combined with each operator’s own network unit costs to give a blended economic cost of traffic.

2. Calculate the standalone economic costs of each operator on the basis that it incurs the total expenditures for the total available network to the operator, but carries its own volume of traffic. This gives the standalone unit costs of traffic on the shared network, and can be used to assess the magnitude of economies of scale achieved from infrastructure sharing. It is important to note that the standalone unit cost of traffic is not the appropriate regulated pricing basis for mobile termination.


303-46

3. Examine the expenditure contribution or inter-payment mechanisms (if any) for each of the parties sharing the network, in order to assess to what extent the economic costs of the shared network are equally borne by each party – to determine whether any higher/lower costs might be outside of the control of the operators.

Recommendation 6: The upgraded model should calculate the shared economic costs of each part of network infrastructure that is shared, which can be blended with each operator’s own network economic costs. The individual circumstances of sharing will need to be calculated separately in the model for the two mobile operator pairings, due to coverage and licence differences. The approach to blending costs will be to take the weighted average of the mobile terminated volumes of an operator that are carried by (where applicable): 1. own GSM network, 2. own UMTS network, 3. shared UMTS network, 4. shared GSM network.


303-46

5 Conceptual issues relating to services

This section discusses the following issues:


Service set Voice and data No Standalone network or vertically integrated Full business costs No

Table 5.1: Service conceptual issues in the original model [Source: Analysys]

The principal requirement of the original model was to understand the costs of voice services – in particular access, call origination and call termination. Therefore, although data volumes were included (i.e. SMS and GPRS), there was no particular need to calculate the unit costs of any other services. However, the model as actually built was in fact able to cost:

• access cost per average subscriber per year • incoming minutes per average minute from other licensed operators (OLOs) • outgoing minutes per average minute on the same mobile network (i.e. on-net minutes) • outgoing minutes per average minute to OLOs • SMS per average originated message • GPRS per average Mbyte.

The original model reflected the full business costs of the operators – their network, retail and overheads costs – where wholesale mobile termination costs reflected network costs plus a share of overheads. However, PTS’s original recommendation – to just model network plus overhead costs – was broadened in response to the operators’ requests to include non-network costs. This specifically required operators to submit cost data on their retail activities, including sales, marketing, advertising, handset/SIM subsidies, other acquisition and retention costs, billing, customer care and retail staff.

By including UMTS technology, additional network services may be required in the model to capture the permutations of voice and data services that occur with two technology generations. However, consistent with the views submitted by some industry parties, it is recognised that at a retail level, GSM and UMTS subscriptions are generally (from a voice perspective) indistinguishable – therefore the additional UMTS services proposed for the model will be considered specifically as network services which can be blended together to provide an over-arching retail or wholesale service such as mobile termination. An indicative revised list of network services is given below in Table 5.2.


303-46

Service to be costed

Access per average GSM handset per year

Access per average UMTS handset per year

Incoming minutes to GSM network, from OLOs

Incoming minutes to UMTS network, from OLOs

GSM on-net minutes

UMTS on-net minutes

Outgoing minutes from GSM network , to OLOs

Outgoing minutes from UMTS network, to OLOs

GSM network SMS per average originated message

UMTS network SMS per average originated message

MMS per average originated message

GPRS data per average Mbyte

UMTS data per average Mbyte

Table 5.2: Indicative

revised list of

services for the

upgraded model

[Source: Analysys ]

There will also be non-network costs for UMTS services, but since modelling these would require additional information to be submitted on retail activities, it is proposed that the extension of the model is focussed on the W–CDMA network aspects and business overhead costs, combined with an exogenous estimate of UMTS retail economic costs.

It is proposed that this exogenous estimate of UMTS retail costs should be based on the retail cost calculations in the original model as well as any top-down cost data on retail activities submitted by the mobile operators in response to the data request. This exogenous estimate will then be used to allocate or mark-up business overhead costs to some or all network and retail activities. The approach to treating business overhead costs will be to examine the cost data submitted by the mobile operators and identify two categories of business overheads:

• indirect costs that vary with the size or activity of network or retail services (such as the HR department, which may vary according to the headcount of the business)

• true common costs which are necessarily incurred irrespective of the size or existence of network or retail activities (for example, the CEO).

Indirect costs will be marked-up to the relevant direct network and retail elements, whereas true common costs will be treated with the mark-up mechanism proposed later in this document.

Recommendation 7: The upgraded model should reflect the corresponding voice and data (Release 99 and HSPA) services which exist on the UMTS network. Wholesale termination costs should be calculated according to blended network costs, plus a share of business overhead common costs. This share of business overhead common costs should be based on identifying indirect costs relevant to network activities, and common costs that are shared across all parts of the business. Accordingly, explicit modelling of UMTS retail costs will not be necessary in the upgraded model.


303-46

6 Conceptual issues relating to implementation

This section discusses the following issues:


WACC Use Capital Asset Pricing Model (CAPM) No

Depreciation method Economic depreciation No Increments Two; traffic and subscribers Yes Type of increments Average No Years to calculate results for All past and future years Yes Mark-up mechanism Equi-proportionate mark-up (EPMU) No Inclusion of network externality No No

Table 6.1: Implementation conceptual issues in the original model [Source: Analysys]

6.1 WACC

The original model used the Capital Asset Pricing Model to derive a value for the nominal, pre-tax weighted-average cost of capital (WACC). The WACC was derived

• in nominal terms, since interconnection prices are fixed in nominal terms • pre-tax, since regulatory price-setting normally uses a pre-tax cost of capital to apply to the

capital base to calculate annual capital costs before taxes.

The value to be used within the upgraded model may be considered in separate work by PTS and therefore falls outside of the scope of the present Paper.

6.2 Depreciation method

Economic depreciation was employed within the original model as best practice, since it

• allows all costs incurred over the lifetime to be considered • allows operators’ expenditures which occurred in the past to be reflected in present and future

charges, to the extent that they are still being (or can be) recovered.

This approach is forward-looking (through the application of modern-equivalent asset (MEA) price trends) and takes into account the full time series of operation. It also requires forecasts of future equipment prices and demand. This methodology should continue to be used, for consistency with the original model.


303-46

6.3 Mark-up mechanism and network externalities

The original model was designed to use the equi-proportionate mark-up (EPMU) mechanism to mark up common costs for services. This was because, although Ramsey Pricing and Network Externalities were discussed as potential options, PTS rejected their application. The upgraded model should therefore apply the same approved principles, using an EPMU mechanism for common costs, and excluding network externality surcharges.

6.4 Increments

An incremental cost is that incurred in supporting a particular increment of demand, assuming that other increments of demand remain unchanged. The incremental cost can also be calculated as the avoidable costs of not supporting the increment. There is considerable flexibility in the definition of the increment – or increments – to apply in a costing model, and the choice should be suitable for the specific application. Possible increment definitions, as illustrated schematically in Figure 6.1 below, include the following:

• marginal unit of demand for a service • total demand for a service • total demand for a group of services • total demand for all services in aggregate.

A B C D E

e.g. fixed annual spectrum fee

e.g. Chief Executive

Variable cost

Attributable fixed cost

Shared cost

Common cost

Service

Marginal unit of demand Total demand for a service

Total demand for a group of services Total demand for all services

Figure 6.1:

Possible increment

definitions [Source:

Analysys]


303-46

The application of large increments means that any economies of scale that are generated across the large scale of the business are incorporated into the incremental cost. This therefore maximises the incremental cost and minimises the common cost, as well as requiring that the large incremental cost is shared out amongst the constituent services in some way.

Also, the adoption of a large increment – in the case of a mobile network, “traffic” – means that all the services that are supplied are treated together and equally. In cases where one of those services is regulated (as in the current situation), the regulated service neither bears, nor benefits excessively from, the higher (or lower) costs arising from economies of scale. Applying a large traffic increment also confines common costs to an assessment of the minimum network deployment required to offer service – in the case of the original model, this was set according to those network assets which were invariant with demand in the long run.

The increments chosen in the original model were average increments for GSM traffic and subscribers. Average increments were used since this produces the most workable model structure, and is an approach that is frequently used in regulatory costing. Average increments also have the property of sharing costs between services in a (proportionately) equal way. In the upgraded model, the definition of GSM incremental and GSM common network elements will need to be expanded to cover:

• dedicated GSM assets (e.g. a TRX) • dedicated UMTS assets (e.g. a NodeB) • shared assets (e.g. a radio site).

Therefore, in order to be consistent with the original model, it is proposed that the incremental costs of traffic and subscribers should be expanded to reflect these three types of shared/dedicated network elements, measured across traffic and subscribers with common costs assessed according to the shared/dedicated network elements which are invariant with demand in the long run. This is illustrated in Figure 6.2.


303-46

Dedicated GSM assets

Dedicated GSM assets - common

Dedicated UMTS assets - common

Shared assets

Shared assets - common

Business overhead common costs

Incremental

Applicable to GSM only services

Dedicated UMTS assets

Incremental

Applicable to UMTS services

Incremental

Applicable to GSM and UMTS services

Retail incremental

and common costs

Dedicated GSM assets

Dedicated GSM assets - common

Dedicated UMTS assets - common

Shared assets

Shared assets - common

Business overhead common costs

Incremental

Applicable to GSM only services

Dedicated UMTS assets

Incremental

Applicable to UMTS services

Incremental

Applicable to GSM and UMTS services

Retail incremental

and common costs

Figure 6.2:

Proposed increment

definition [Source:

Analysys]

Recommendation 8: Apply average traffic and subscriber incremental costs to the upgraded model, expanding the cost categorisation to reflect dedicated GSM or UMTS assets and shared assets for the purposes of calculating the underlying costs of network services. Common costs would be assessed according to shared and dedicated network elements which are invariant with demand in the long run.

6.5 Years to consider

The original model considered the period 1990–2009. This extended model period improved the model’s capability to reconcile volumes and costs with the past and therefore, potentially, the present and future. A full time series allowed the consideration of all costs over time, providing the greatest clarity within the model as to the implications of adopting economic depreciation.

The upgraded model will be developed during late 2007 and early 2008, and the time series should be extended past 2009 to enable PTS to use the model to inform termination pricing for several years into the future. There are several options as to how far to extend the model in time:

• as far as PTS would like to continue to use it for the purposes of the price regulation of mobile termination (e.g. 2010, 2011 or 2012)

• to the end of the expected lifetime of GSM networks (somewhere between 2015 and 2020)

• to the end of the expected lifetime of UMTS networks (between 2020 and 2030).

Given that the upgraded model will include both GSM and UMTS technologies, it will be of most use if it covers the full lifetimes of all the networks that it is considering. GSM networks are likely to have a lifetime of 20 years or more, and the same can be expected of UMTS. Therefore, modelling up to 2029 (a 40-year period from the model’s starting point of 1990) should allow all


303-46

costs associated with both GSM and UMTS networks to be captured. The upgraded model will then be able to:

• estimate windfall losses/gains arising from a change of cost paths, and provide greater clarity as to the recovery of all costs incurred from services

• provide the widest range of information with which to understand how the costs of the operators vary over time and in response to changes in demand or network evolution

• include additional forms of depreciation (such as accounting depreciation) with minimal effort.

Recommendation 9: The period of the upgraded model should be extended from 2009 to 2029 to capture the full evolution of mobile networks from GSM to UMTS.


303-46

7 Model specification

This section describes features of the model design specific to the proposed upgrade of the network design algorithms.

After the original model was completed in 2004, two quality checks were carried out, in 2005 and 2007. In both cases, Analysys examined the model in order to assess the quality and level of its projections for the cost of mobile termination. A data request was sent to the three operators modelled (TeliaSonera, Vodafone/Telenor and Tele2) to request key data, and use was also made of PTS’s latest market information. TeliaSonera provided responses on both occasions, covering most of the requested data points, but on both occasions Tele2 and Vodafone/Telenor declined or were not able to provide additional data. The quality-checking process was carried out in four stages:

• Stage 1: Compare the original modelled market volumes with the latest PTS market data • Stage 2: Compare the original modelled deployment of network assets against the latest actual

data submitted (this was only possible for TeliaSonera) • Stage 3: Compare the original modelled capex and opex with the latest submitted data • Stage 4: Assess the combined effect of the above on the final service costs calculated within

the original model.

In the 2005 quality check, no adjustment was made to the demand volumes or network asset calibration in the model. The model was found to underestimate TeliaSonera’s demand and certain network elements. However, the magnitudes of these underestimates were relatively small and would be mitigated to an extent by adjusting demand upwards and applying effects which were not present in the original model, such as dedicated GPRS channels. In addition, despite underestimating network element volumes, the original model appeared to:

• overestimate network opex • underestimate network capex.

In the 2007 quality check, the model was updated with actual market volumes for 2003–06 using the latest data provided by PTS. However, it was not re-calibrated. This quality check found that, in general, the original model significantly underestimated GSM demand. When the actual (higher) traffic volumes were incorporated into the model, it was found that:

• a larger BTS network was deployed, incurring greater costs (both capex and opex), but this larger network appeared to be aligned reasonably well with actual networks

• total capex over the period 2003–06 was close to the actual value for the period • total opex over the same period was overestimated.

The two quality checks therefore demonstrated that the network design algorithms for the GSM network in the original model were reasonable. The primary issue for the upgraded model is the algorithms that will need to be implemented to incorporate W–CDMA technology. Particular


303-46

consideration will need to be given to a large number of network design aspects, discussed in the sections below. These design aspects relate to the following areas:

• W–CDMA network loading • coverage parameters • equipment capacities • projection of volumes in the mobile networks • network topology.

7.1 Network loading

Call attempts and unsuccessful calls

The number of successful calls is related to the traffic volume according to the average duration of a successful call. This may vary by call type (incoming, outgoing, on-net). For each successful call, there will be a number of additional unsuccessful calls – defined as being instances where the call is not answered by either the subscriber or voicemail, and results in no conveyed minutes.

Busy-hour profile Most mature European mobile networks exhibit approximately 250 busy days per year (normal working days). These busy week-days correspond to around 80% of annual traffic, and approximately 8% of each busy day’s traffic occurs in the busy hour. Data submitted by the operators should assist in identifying the relevant statistics for the Swedish mobile networks, if possible separately for GSM and UMTS. This may also reveal the relevant parameters for non-voice services, such as videotelephony, SMS, MMS and packet data.

The busy-hour parameters will be used to convert annual demand into a busy-hour load (in Erlangs).

The SMS and packet data busy-hour profile is expected to be different from the profile of voice.

Treatment of UMTS packet data services

Packet data services carried over W–CDMA will be modelled according to the radio network load they present to the cells – and therefore the number of channels per bearer type will be estimated. The following parameters are proposed:

• 1 channel element (CE) for a voice call, 12.2kbit/s • 4 CE for 64kbit/s packet data • 8 CE for 128kbit/s packet data • 16 CE for 384kbit/s packet data • 32 CE for five HSPA codes.

In addition, we will include eight common control and signalling CE per sector. It will be important to test at an early stage whether data traffic


303-46

currently accounts for a non-negligible volume of network traffic, or whether the capacity dimensioning of the network is entirely voice-dominated.

Location updates and handovers

In the W–CDMA networks, additional Erlangs are generated by soft-handover traffic (approx 20%) for circuit switched and R99 packet communication. In addition, a number of channel elements (per sector) are dedicated to common control and signalling functions. Soft handover does not exist for HSPA traffic because users are handled by fast-scheduled channels.

Blocking probabilities

Using the data submitted by operators on the cell upgrade blocking probability limit, it should be possible to determine whether the blocking probability of the air interface and the core network differ for GSM and UMTS operations.

7.2 Coverage parameters

Geotypes Three geotypes (urban, suburban and rural) were defined in the original model, and these will be retained in the upgraded model. Operators should submit data that allows the calculation of traffic load and network elements on a geotype basis. The geotypes are defined as follows:

• Urban: 2109 km2, 4 977 620 population • Suburban: 23 780 km2, 3 150 522 population • Rural: 431 473 km2, 722 206 population.

Cell radii Cell area and cell radius govern the amount of area covered by a site. As in the GSM case, the hexagon-per-BTS model will be used, assuming a cell pi of 2.6, based on the hexagon-per-site model.4 The cell radius for coverage will vary by geotype and frequency. The cell radius in a UMTS network also varies by expected (or actual) network load due to the effect of cell breathing. The model will therefore be required to reflect an assumed cell load (either implicitly or explicitly).

4 The area of a circle is πr2; the area of a hexagon is 2.6r2. In the situation where an operator submits information on a hexagon-per-

cell basis, we will perform the necessary transformations to ensure correct application of the coverage model.


303-46

Spectrum differences

Spectrum differences are expected to be applied to coverage as follows:

• choice of primary spectrum5 used to achieve coverage (900MHz, 1800MHz and 2100MHz)

• cell radius for 900MHz, 1800MHz and 2100MHz coverage sites.

GSM and UMTS spectrum will be treated separately. There will be no consideration of secondary spectrum used for UMTS (it is all primary 2100MHz spectrum). It is anticipated that, within each geotype, the cell radius for each technology will obey the following relationships:

• 2100MHz cell radius < 1800MHz cell radius • 1800MHz cell radius < 900MHz cell radius.

Cell types The original model included micro cells and pico (indoor) cells for GSM deployments. If operators indicate that they use equivalent equipment in their W–CDMA networks, then UMTS micro cells and UMTS pico (indoor) cells will be incorporated into the model.

NodeB sectorisation

The original model assumed that GSM macro cells had on average two sectors, with micro cells and pico cells being omni-sectored. Unless data from operators indicates otherwise, it will be assumed that all W–CDMA macro cells (NodeBs) are tri-sectored. Similarly, if UMTS micro cells and pico cells are used, they will be assumed to be omni-sectored.

Traffic by geotype Following an approach similar to that used in the GSM model, it is proposed that the evolution of UMTS traffic by geotype across Sweden should be modelled according to the demographic coverage roll-out of the W–CDMA network. There will be an underlying traffic density factor, consistent across all operators. Its value will be highest in urban areas and lowest in rural areas. Information on the long-term urbanisation trends in Sweden may be considered as part of the traffic by geotype calculation.

Site sharing Based on data submitted by the operators, the model will apply a factor representing the proportion of an operator’s existing GSM sites which are available for the deployment of W–CDMA NodeBs (for a UMTS-only operator this will be 0%). Therefore, in each year, there will be a combination of NodeBs erected on new sites and on existing sites. The degree of site sharing will also need to take into account the alternatives of “umbrella” versus “street level” deployments.

Infrastructure For a discussion of the treatment of the shared infrastructure in Sweden, see Section 4.3. In the situation where equipment in the network is no

5 Primary spectrum is the frequencies used for deploying national coverage.


303-46

sharing longer required (e.g. in order to meet GSM demand) and is also not required to support (or cannot support) UMTS services, then it is appropriate to consider whether the decommissioning of such elements implies additional decommissioning costs, potential revenue from the sale of a working asset, or other decommissioning issues.

7.3 Equipment capacities

Radio element capacities

It will be assumed that 1 × 5MHz paired carrier = up to 80 (or alternative) channel elements (CE) per sector, in 16-channel kits or 64-channel kits (CK). Channel elements will be pooled at the NodeB, however.

Backhaul links It is proposed that E1 backhaul links will be modelled with a capacity of 80 circuits (24kbit/s channel rate). The average amount of backhaul capacity needed per UMTS site will determine whether the site requires 1, 4, 8, 16 (or possibly more) E1 units of backhaul capacity. A proportion of this backhaul will be provided by leased lines, the remainder by microwave links (of a certain number of hops per link).

RNC switches RNC switches will be assumed to have a capacity in terms of NodeB and throughput supported (CS Erlangs plus PS Mbit/s). E1 ports into and out of the RNC will be modelled in the same way as for BSC switches in the original model.

RNC–UMTS MSC links

Links from RNC to UMTS MSC units can be provided by three methods: remote microwave link (2–32Mbit/s or higher), leased lines (2–32Mbit/s or higher) or co-located in-building links (2–32Mbit/s or higher). These will be dimensioned at 120 circuits, or 2Mbit/s, per E1.

UMTS MSC (including VLR functionality)

UMTS MSCs will be deployed according to the CPU processing demands of the network generated by a number of services and processes, including the following:

• subscriber authentication • incoming and outgoing circuit switched call set-ups • SMS message send and delivery • subscriber location updating.

This will be captured using busy-hour loading and will also factor in a minimum number of locations (defined by the network topology).

In addition, the number of UMTS MSCs should be sufficient to support port demands according to the ports


303-46

• to the RNC • to other UMTS MSCs • to PoI • to other servers (HLR, SMSC, etc.).

Network software Network software is added to the network as either periodic releases or as software for new features. It is proposed that network software is modelled per UMTS MSC switch and per NodeB, depending on the information provided by the operators.

UMTS MSC–UMTS MSC links

The proportion of traffic which is carried on inter-switch links will be used to dimension a fully-meshed inter-MSC network. Circuit- and packet-switched traffic will be dimensioned on multiple E1 links per inter-switch route. The number of ports associated with these links will also be modelled.

Data servers The model will include HLR, SMSC, VMS, GGSN and SGSN switches serving both GSM and UMTS traffic loads.

Capacity utilisation factors

It is proposed to that a set of maximum utilisation factors will be applied to the modelled network element capacities, to represent a reasonable average maximum utilisation. For example:

• Switches are not utilised at 100% capacity since there will be transient variations in demand which would take utilisation above 100%. 80% maximum utilisation would be the proposed starting assumption.

• The geographic location of demand (particularly in the radio network) cannot be predicted at any point in time and varies according to the specific location and motion of subscribers. As such, it is inevitable that radio capacity installed on a geographic basis cannot be 100% utilised at all times.

• The availability of suitable sites for deployment cannot be guaranteed: for example, radio sites cannot be ideally located to achieve a perfectly tessellating hexagonal coverage map because the availability of sites is dictated by buildings, clutter and land usage in each district. In such circumstances a realistic but less-than-perfect deployment may be necessary – this is normally described as scorched-node calibration.

Analysys will explore to what extent it is possible and accurate to separately identify these capacity utilisation effects.


303-46

7.4 Demand inputs and projections

The upgraded model will require additional data and forecasts on market share, SIM type and traffic volumes. These are discussed in turn below.

7.4.1 Market share

Market share can be quoted in terms of subscribers, traffic, or even revenues. However, for the purposes of developing a market projection, it is proposed that the model calculates the four network operators’ shares of the registered SIM base (including their hosted service providers and MVNOs, but excluding the two niche sub-national operators). If the submitted information on SIM (in)activity allows, then it may be possible to understand whether specific operators have different inactive proportions of their market share base – and to make market share projections accordingly. It is proposed that the year-end market share is calculated.

7.4.2 SIM type

Within Sweden, a number of distinct registered SIM types are likely to exist, although from a retail perspective subscribers may not be aware of the difference in SIM type, or the two networks which carry their traffic. SIM types are likely to include:

• GSM • UMTS • GSM data-only SIM • HSPA data-only SIM.

The number of data-only SIM cards and the proportions of SIM cards that are GSM or UMTS will vary over time (i.e. due to migration). These SIM types are likely to be modelled in an operator-specific manner, again utilising a year-end calculation. Overall market ‘data-only SIM’ and ‘GSM-UMTS migration’ proportions will be calculable.

7.4.3 Traffic volumes

The previous steps, discussed above, calculate a year-end number of registered SIMs for each operator, by SIM type. The next step is to calculate the total amount of volume for each retail (or wholesale) service. This usage can be mapped to each network type over time, for historical time periods (where information allows). For future years this usage by network will be projected. The following traffic categories will be used:


303-46

Volume measure Network type

GSM UMTS GSM data-only HSPA data-only

Voice

SMS

GSM packet data (GPRS)

UMTS packet data (Release-99) (HSPA)

UMTS packet video

Table 7.1: Traffic by network type [Source: Analysys]

The following traffic categories per network are proposed, subject to submitted data or available estimates:

• voice: incoming from other operator, outgoing to other operator, on-net • SMS: incoming, outgoing, on-net • GSM (GPRS) packet data: Mbytes • UMTS packet data: Mbytes by data rate – 64kbit/s, 128kbit/s, 384kbit/s • HSPA packet data: Mbytes by data rate – 1.8Mbit/s, 3.6Mbit/s, 7.2Mbit/s (category 12, 6 and

8 respectively).

7.4.4 Projections

It is proposed that, subject to availability of submitted data, historical demand parameters will be calculated up to the end of the calendar year 2006. For 2007 onwards, market shares and traffic volumes will be projected according to a converging evolution in market share and volumes based on Analysys’s most likely forecast view. This forecast convergence is consistent with the approach taken in the original model.

It is also important to understand the impact of a number of other projections on the results – which will be examined using scenarios in the model:

• impact of Nordisk Mobiltelefon on voice and data volumes of the four main operators • volume of HSPA traffic • availability of additional spectrum for UMTS/HSPA at 2600MHz.


303-46

7.5 Network topology

NodeB

RNC

PCU

3G MSC

GGSN SGSN

3G MSC

HLR

NMC

CS POI

PS Internet

E1

SMSC

E1

CKCK

CKCK

CKCK

CKCK

CKCK

CKCK

E1

NodeBNodeB

CKCK

CKCK

VMS

E1

b ca

d e

f

g

h

ij

k

lm

n

Figure 7.1:

UMTS mobile network

elements [Source:

Analysys]

See text below for

explanation of items

(a) to (n)

A generic network design will be developed which can accommodate the different network topologies of the Swedish mobile operators. The proposed logical structure of the network is shown in Figure 7.1. It consists of the following network elements:

a. Tri-sectored UMTS-macro site. In the case of UMTS, the primary spectrum is 2.1GHz. Three-sectored sites are predominantly deployed in Sweden, because of the standardisation of network kit and improved coverage compared to omni-sectored sites. The minimum number of carriers per sector (one to three, depending on configuration) would be deployed in order to provide the required capacity.

b. Omni-sectored UMTS micro site

c. Omni-sectored UMTS pico (indoor) site

d. Microwave backhaul (2-32Mbit/s)

e. E1 leased-line backhaul (2-32Mbit/s)

f. E1 incoming ports to RNC


303-46

g. RNC switch

h. Packet control unit for UMTS packet traffic: the PCU grooms the packet traffic at the RNC and sends it to the logically separate packet-switched transmission

i. E1 incoming ports to the UMTS MSC

j. A fully-meshed network connecting the UMTS MSCs

k. Circuit-switched point of interconnection with other licensed operators

l. Packet data switches: subscriber GPRS serving node (SGSN) and gateway GPRS serving node (GGSN)

m. Packet-switched connection to Internet

n. Network servers and management systems, which perform a wide variety of supplementary network, management and authentication functions. These include SMS centre (SMSC), network management centre (NMC), home location register (HLR) and voicemail system (VMS).

Recommendation 10: The upgraded model should include network design algorithms specifically tailored to a W–CDMA network.


303-46

Annex A: Glossary of technical abbreviations

Acronym Description Acronym Description

2G Second generation of mobile telephony NMC Network management centre

3G Third generation of mobile telephony NMT Nordic mobile telephone system

BSC Base station controller PCU Packet control unit

BTS Base transmitter station or base station POI Point of interconnect

CAPM Capital asset pricing model RNC Radio network controller

CCA Current cost accounting SGSN Subscriber GPRS serving node

CE Channel element SIM Subscriber interface module

CK Channel kit SMS Short message service

E1 2Mbit/s unit of capacity SMSC SMS centre

EPMU Equi-proportionate mark-up SGSN Subscriber GPRS serving node

GGSN Gateway GPRS serving node SIM Subscriber interface module

GPRS General packet radio system SMS Short message service

GSM Global system for mobile communications SMSC SMS centre

HCA Historical cost accounting TRX Transceiver unit

HLR Home location register UMTS Universal mobile telecommunications systems

LRIC Long-run incremental cost VLR Visitor location register

MEA Modern equivalent asset VMS Voicemail system

MSC Mobile switching centre WACC Weighted average cost of capital

MVNO Mobile virtual network operator W–CDMA Wideband code-division multiple access


303-46

Annex B: Revisions to approach following industry consultation

• Recommendation 1 has been expanded to include MVNOs supported by each mobile operator.

• Recommendation 2 has been expanded to consider the cost impact of uncertainties regarding 900MHz spectrum that the mobile operators are facing.

• Recommendation 3 has been expanded to explicitly refer to the multi-technology situation facing the incumbent mobile network operators, distinct from Hi3G.

• Recommendation 4 has been expanded to include investigation of the voice costs which may arise because of the uncertainties in non-voice (particularly HSPA) traffic.

• Recommendation 5 has been extended to include investigation of the costs under different circumstances for the long-term status of GSM and UMTS networks.

• Figure 4.4 has been revised to illustrate that Hi3G’s national roaming with Telenor is not confined to remote parts of the country, but instead occurs throughout the Telenor network.

• Recommendation 6 has been extended to define the approach to calculating blended costs for each individual network operator.

• Section 5 has been revised to clarify that the model will identify network service costs, which form the underlying components of non-technology-distinct voice services at a retail level. A revision to the proposed treatment of business overhead costs has also been provided to explain the distinction between indirect and common business overhead costs.

• Recommendation 7 has been updated to reflect the revisions made to Section 5.

• The model specification has been revised as follows: – channelisation of UMTS packet services – soft-handover for R99 packet services – reflection of cell breathing – consideration of urbanisation trends – deployment of “umbrella” or “street-level” UMTS networks – potential issues associated with equipment no longer required in the networks – radio equipment capacities – network software per NodeB – scorched-node calibration and capacity utilisation factors – reflecting the characterisation of total demand at a retail (or wholesale level) divided

across network types, rather than according to subscriber SIM types.