Transcript The LRIC model of UK mobile network costs, developed for

[30/1/02 -02]
The LRIC model of UK mobile network costs,
developed for Oftel by Analysys, September 2001
A Manual for the Oftel model
Working paper for Oftel, 29 January 2001
2
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
Executive summary
3
Executive summary

This working paper presents a comprehensive description of the long-run incremental cost
(LRIC) model of UK mobile network costs, developed for the UK regulator, Oftel, by
Analysys, during 2001

The model was made available by Oftel in conjunction with its statement on mobile
termination in the UK, and can be downloaded from the Analysys Web site. Although this
model is freely available, it is copyright of the UK Crown, and should not be used for any
purpose other than the review into mobile termination in the UK

this document does not contain details of the Excel-related mechanics of the model

instead, it provides details of the theory that underlies the model, and details of the nature
of calculations employed (but not their Excel implementation)
4
Executive summary
Related documents

The LRIC model of UK mobile network costs, developed for Oftel by Analysys


download from www.analysys.com
Oftel statements related to the review of mobile termination in the UK

download from www.oftel.gov.uk
5
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
6
Introduction
LRIC modelling is a method of calculating costs which employs a specific set
of costing principles

Long-run incremental cost modelling relates to:

a consideration of costs over the economic lifetime of assets (long-run)

the attribution of costs to specific services

Estimates the economic costs of installing, maintaining and operating a mobile network

Estimates the cost to a new entrant of providing the same service as the existing network operator

Identifies the structure of costs – how they vary with the level of demand and range of service offerings

Advantages are:

a good predictor of volume/cost movements

represents an economically rational approach to pricing cost-based services over time

of increasing interest to regulators, especially for validation of interconnect arrangements, because
cost-orientated

of paramount interest to new entrants

forward-looking
Introduction
7
LRIC cost modelling is supported by major regulators and other organisations

Supported by the FCC, EC and IRG (Independent Regulators Group) for costing mobile
termination

Applied by OFTEL in its current proposals for the regulation of mobile termination rates in the
UK
8
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
9
Background
Oftel was required to consider a range of issues
when setting interconnect prices
Prices
Pricing method
e.g. LRIC, LRIC+
Costs
Other factors
e.g. externalities
Costing method
e.g. forward-looking
economic costs
Data
e.g. unit
input costs
Assumptions
e.g. demand
forecasts
Background
10
The models developed for Oftel by Analysys only derived the costs of mobile
termination, and enabled a number of mark-up regimes to be applied
Illustration of
alternatives
not policy
Prices
Pricing method
e.g. LRIC, LRIC+
Costs
Other factors
e.g. externalities
Costing method
e.g. forward-looking
economic costs
Data
e.g. unit
input costs
Assumptions
e.g. demand
forecasts
11
Background
Analysys constructed the 1998 and 2001 LRIC models for Oftel

In 1998, Analysys constructed a bottom-up LRIC model for Oftel, to assist Oftel in its 1998 review of the
price of calls to mobiles

This model calculated the costs of :


a reasonably efficient new entrant

in a (hypothetically) fully contestable market

with the demand parameters of either Vodafone (GSM 900) or Orange (GSM 1800)

for (year average of) the financial year 98/99
In 2001, Analysys began the process of updating this model to reflect the needs of Oftel for the next review
of the price of calls to mobiles (completed September 2001)
12
Background
A number of areas of the model were highlighted for improvement

Enable the model to calculate costs for the years 2000/01–2005/06

Improve specific areas of the model:


update and refine data and assumptions in the model, with the co-operation of the UK
operators

review methodological issues, with input from operators, to improve the accuracy and
suitability of the network deployment algorithms

make the model algorithms and calculations more explicit
Update the model to reflect the current and expected development of the mobile market:

current: SMS, emerging HSCSD and GPRS services, increased expectations of the
“quality of mobile network coverage”

expected: increased take-up of data services (HSCSD, GPRS and latterly UMTS),
eventual decline in SMS in favour of packet based messaging services, simultaneous
operation of UMTS voice and data networks by the four UK operators
13
Background
In order to calculate costs out to 2005/06, forecasts of the UK mobile market
and associated network deployments were required


It was important to establish consistent forecasts, calculations and model algorithms e.g.

the allowance for growth assumed in deploying the network was consistent with the
growth in market demand

that the nature of the (hypothetical) competitive market was correctly and consistently
represented
Taking into account the (2000/01 real terms) model results, Oftel derived P2000/01, P2005/06
and X


these parameters (P = price; X = percentage price decline) were important in setting the
regulated price cap
The UK mobile market was forecasted in
terms of:

Network deployment forecasts required
time series for:

subscribers


minutes of use (incoming, outgoing,
on-net)
demand drivers (e.g. busy hour traffic
proportions)

data service take-up (subscribers,
technologies, megabytes of use)
network design parameters (e.g.
traffic by cell type)

equipment unit costs

14
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
15
The model
The costs calculated by the model developed by Analysys represent a unique
implementation of LRIC theory and regulatory policy …


The model developed by Analysys in 1998 calculated the long run costs of:

a reasonably efficient new entrant

in a (hypothetically) fully contestable market*

with the demand parameters of either Vodafone (GSM 900) or Orange (GSM 1800)

for (year average of) the financial year 98/99
However, the model developed by Analysys in 2001 calculated the critically different long run costs of:

a reasonably efficient operator that launched service in 1992/93 (corresponding with the launch of
GSM in the UK)*

in a market with the assumed level of contestability*

with 25% share of the total mobile market from 1992/93 to 2002/03, declining to 20% share of the
total mobile market by the end of 2009 (corresponding with the entry of the fifth player to the UK
market)

for the (year average) of financial years 2000/01 to 2005/06
* See later section on Economic Depreciation for definition of these terms and justification of approach adopted
16
The model
… but the cost modelling is still based on sound techno-economic principles


Bottom-up
A ‘scorched-node’ approach was adopted, so that the network design reflects the actual number of base
stations and switch sites currently deployed

a scorched-node deployment is one that evolves over time and is constrained by the history of
deployments

conversely, a scorched-earth deployment is one which has no historic constraints, and can be deployed
in an optimal fashion

Modern technologies (for example, those currently being deployed) are used throughout (MEAs; modern
equivalent assets)

Sufficient capacity to meet present (coverage and demand) requirements is provided; plus an allowance for
reasonable future growth, but no more

Incorporating (a variant of) economic depreciation for calculating economic costs

Deriving the long run average incremental costs:

average costs are calculated rather than marginal
17
The model
Scorched node approach
Background to the scorched node approach

Networks develop over time in response to changes in demand (or forecast demand)


The location of network nodes is dictated to at least a degree by the availability of suitable sites
on the ground


As a result of this evolutionary development, networks are rarely truly optimal for current
(or currently forecast) market conditions
Such sites are rarely in the ideal location from a theoretical perspective – another reason
for networks being less than optimal
Radio network design is a complex process, involving a very large number factors and design
parameters, not all of which are measurable in advance

To accurately capture every nuance of these algorithms in a predictive cost model would
be excessive (and almost certainly impossible given the reliance to some extent on
information that can only be measured once the network is in place)
18
The model
Scorched node approach
The rationale for the scorched node approach



The scorched node approach accepts that:

these are real processes that increase the cost of providing services, and

that it is impossible to accurately capture the impact of such highly complex processes as these in a
purely predictive model.
The scorched node approach therefore relies instead upon actual statistics about the design of operators’
networks as predictors of the aggregate impact that these effects would be likely to have on the network
design of an operator, including that of a new entrant.
NB Not because incumbents’ have to continue operating their existing networks:



If the market were contestable (even if not fully contestable) then incumbents’ would have to set prices
in line with those that a new entrant would charge;
New entrants would not have to recreate the existing design of an incumbent’s network if that were
less than fully efficient, but they could be expected to suffer the same problems as incumbents already
have, when rolling out their networks.
NB This does not mean that the modelled operator has to have exactly the same number and distribution of
nodes as does a real operator, merely that the relationship between the drivers of node deployment and actual
node placement, are similar in the model to those actually seen in the networks of real operators.
19
The model
Scorched node approach
Notes re implementation in the
LRIC Model of UK Mobile Network Costs

Information about the networks of the four UK mobile network operators was collected from a variety of
sources – in particular the number of base stations, BSCs and MSCs

Information about the coverage and traffic carried by each of the networks was also obtained or estimated

The network design algorithms and parameters in the model were then fixed at reasonable values (based on
general industry data)

A specific parameter of the network design algorithms (the “scorched node utilisation”) was then adjusted
for each network element until the number of units of that element predicted by the model was reasonably
close, for all network operators, to the actual number of units of that element believed to be in use in the real
networks


The resulting value for the scorched node utilisation parameter simply describes how much lower (or
higher) than expected (on the basis of the standard network design algorithms and parameters used in
the model) the actual utilisation of network elements really is
The model can then predict the number of nodes that a 25% market share operator would be likely to have,
with a reasonable degree of accuracy, based on the actual number of nodes in use by the UK operators today
20
The model
The scope and detail of the model is critical

The model aims to capture:

all relevant network elements and business activities

all relevant expenditures:
– capital investment
– operating expenses
– return on capital employed

The level of detail in the model should be sufficient:

for the network design to reflect actual industry practice rather than some hypothetical
optimum or simplification

to capture significant factors that influence the total cost of the network, yet should not be
more complex than is absolutely necessary
21
The model
Key inputs fall into five broad categories

Service demand levels

Network design rules and parameters

Equipment unit costs (and price trends)

Cost of capital

Service routeing factors
22
The model
The key outputs are a number of cost figures


For each year, the model outputs:

total common cost

total incremental costs

unitised, un-marked up incremental cost per service

unitised marked-up cost per service, for a number of alternative mark-up regimes
Unitised costs represent:

total costs associated with an increment divided by number of demand units of that
increment
The model
23
Mark-ups

Unmarked-up costs represent the raw incremental cost associated with each increment, without
recovery of common costs

Common costs may be recovered by marking-up some or all of the raw incremental costs of
services – increasing prices of those services to ensure recovery of the costs common to some
or all services

A number of different mark-up regimes are possible – see later for details

In all cases mark-ups are calculated and applied as a percentage increase on raw incremental
costs

The recovery of common costs from services is therefore done by reference to incremental
costs (possibly more or less weighted according to the service) and not by reference to any
common unit of demand or supply (which is typically how such costs would be allocated to
services in a fully allocated cost model)
24
The model
The model flow consists of six major building blocks; information flows from
input, to calculation, to output …
A
Cost drivers,
services and
increments
B
Forecast of
demand
2000–2006
C
E
Network
design
1
3
Network
element
costing
2
D
Economic
cost
4
F
5
Service
costing
25
The model
… which are shown in brief in this section


The following slides indicate the main data, assumptions, calculations and information flows
associated with each of the:

six building blocks identified

five information flows
Following this section, each section of the model is discussed in greater detail
Legend
Data or
assumptions
Information flow
Calculations
or Outputs
Major
elements
26
The model
A. Cost drivers, services and increments
Define what the
drivers of cost are
Define the
associated services
and increments
Define how the
increments
will interact
Cost drivers,
services and
increments
27
The model
B. Forecast of demand 2000–2006
S-curve
penetration
Market
shares
Minutes
per sub
2G/3G
partition
SMS
penetration
SMSs per
user
MByte user
penetration
MByte
per sub
2/2.5/3G
partition
Mobile
subscribers
2G incoming
minutes
2G outgoing
minutes
Forecast of
demand
SMS
volumes
GPRS
users
GPRS
MBytes
HSCSD
MBytes
28
The model
1. Cost drivers and demand forecasts to network design
The drivers of cost
Cost drivers,
services and
increments
Year average mobile subscribers
Year total incoming minutes
Year total outgoing minutes
Year total SMS messages
Forecast of
demand
2000–2006
Year average GPRS users
Year total GPRS Mbytes
Year total HSCSD Mbytes
Select year:
00/01
01/02
02/03…
Network
design
Select
operator:
GSM 900,
GSM 1800
29
The model
C. Network design
Design
parameters
Selected
year
Coverage
Selected
operator
Demand inputs
Coverage
network
design
Full
network
design
Incremental
network
design
30
The model
2. Network design to economic cost
Selected year
Network
design
Selected operator
Out-turn utilisation profiles
Economic cost
31
The model
D. Economic cost
00/01 MEA
capex
Selected year
00/01 MEA
opex
Economic
lifetime
Opex
trends
Annualisation
percentage
Selected operator
Out-turn utilisation profiles
* Calculation performed for each item
Economic
cost
calculation
Capex
trends
32
The model
3. Network design to network element costing
Coverage network deployment
+
Network
design
Incremental network deployment
=
Full network deployment
Network
element
costing
33
The model
4. Economic cost to network element costing
Economic cost
Economic cost for each item
Network
element
costing
The model
34
E. Network element costing
Economic cost for each item
Coverage network deployment
Coverage
network
cost
+
Incremental network deployment
Incremental
network
cost
=
Full network deployment
Full network
cost
35
The model
5. Network element costing to service costing
Average incremental cost of each
network element per unit output
Network
element
costing
Service
costing
Common costs of coverage
36
The model
F. Service costing
Routeing factors
Average incremental cost of each
network element per unit output
Unitised
incremental
cost per
service
Common costs of coverage
Mark-ups
to recover
common
costs
37
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
38
The model
Define what the
drivers of cost are
Define the
associated
services and increments
Cost drivers
We assume four primary cost drivers
Cost
drivers,
services
and
increments
Define how the
increments
will interact


In a mobile network, the primary drivers of cost are:

the level of coverage required, either geographically, or in terms of quality (in-building penetration,
etc.)

the number of customers (subscribers)

the amount of traffic that is carried on the network

the quality of service (QoS) offered to the customers, in terms of blocking or dropping probabilities
In addition, a range of secondary drivers of cost exist, for example:

number of location updates

number of call handovers
39
The model
Cost drivers
Coverage requirements are defined in terms of population and area coverage
100%
90%

Coverage is often quoted in terms of percentage of
population covered (as per licence obligations)

More useful to a mobile network designer is the
geographical area covered (disaggregated by type):
Population


We define a number of area types that effectively capture
the broad range of radio environments in a country. In the
UK, we used:


Area
60%
100%
Suburban
Urban
Rural
Highway
converting population coverage into area
requirements usually requires detailed
demographics
urban, suburban, rural, highway
For example 90% of the population can be covered in
60% of the land area, comprising all urban, all suburban,
part rural and part highway coverage

strictly speaking, no-one lives on a highway, and
such deployments cover rural motorway-side
towns and villages
40
The model
Cost drivers
Notes re implementation in the
LRIC Model of UK Mobile Network Costs

In-building penetration is not explicitly quantified in the model


The scorched node approach ensures that the level of in-building coverage included in the
model is comparable with that typically provided by UK operators
Likewise, the effects of secondary cost drivers, such as the number of location updates and call
handovers, are not explicitly quantified in the model

The values of other network design parameters have been set conservatively to provide
sufficient capacity to deal with these activities
41
The model
Cost drivers
Customer-driven costs are not significant ...
Infrastructure related



Mobile networks do not have substantial
investments tied up in plant dedicated to
individual customers
However, some elements (such as the
maintenance of a HLR about status of customers)
are sensitive to customer volumes
Handsets

In addition, each customer requires a mobile
handset in order to make or receive calls

The cost/subsidy per handset is the only relevant
cost component and in general considered
separately from customer-driven network costs

The amount of costs associated with handsets
may however be taken into account in the markup regime

Similarly, the (year average) number of
subscribers is used to drive handset costs
Hence, the model contains the (year average)
number of subscribers as a driver
42
The model
Cost drivers
... whereas traffic and quality of service are significant cost drivers
Traffic


Principal measures used when dimensioning
network elements are:

busy hour erlangs (busy hour minutes/60)

busy hour call attempts
Levels of cost drivers are calculated separately
for each traffic-related service, based on the
annual amount of traffic


the use of appropriate annual traffic and
busy hour averaging parameters ensures
that the network is also driven by the year
average load
Traffic cost drivers (incoming, outgoing and onnet voice, SMS messages, GPRS and HSCSD
data traffic) are assumed to be parallel (see next
slide for explanation) and hence can be combined
into a single increment called traffic
Quality of service

Quality of service is an important driver of cost

However, inverting the relationship between
quality of service and cost is a complex
transformation, and does not result in a simple
increment that is orthogonal or parallel to others

Hence we do not define a service increment
called quality of service, with X units:

and cannot determine the cost per unit of
quality (whatever unit that may be)

However, the model contains blocking
probabilities as inputs, so can be used to
investigate the variance of other service unit costs
with quality of service

The base case values for these are:

2% blocking on the air interface

0.1% blocking in the core network
43
The model
Cost drivers
What is the significance of orthogonal and parallel services?
For example, two drivers of cost, each with a corresponding service increment:

If the services are orthogonal, then equipment that supports service 1 does not support service
2 and vice versa

no common costs exist (other than the coverage network, if appropriate)
HLR – for customers only

TRX – for traffic only
If the services are parallel, then equipment that supports service 1 partially or entirely supports
service 2, and vice versa

common costs exist between the services, according to the levels of demand and design
algorithms
TRX – for voice traffic

TRX – for GPRS traffic
dedicated costs also occur for each service, where appropriate
GGSN – for GPRS only
44
The model
Cost drivers
Combining service into a single increment simplifies the calculation
requirements

Most services exhibit both parallel and orthogonal behaviour, depending on the particular
equipment class which they are interacting with:


Resolving the common and incremental costs associated with each increment absolutely is a
complex algebraic calculation and a time consuming process:


for example, HLRs are a dedicated resource for customers; however, the MSC processing
requirement of location updates (a customer driven cost) is shared with the MSC
processing requirements of incoming and outgoing call attempts
such a calculation needs to resolve all combinations of common costs and incremental
cost by considering all possible permutations of the increments
Combining services into a single increment for all demand simplifies the model:

orthogonal service costs are resolved without need for complex calculations

parallel service costs are resolved on the basis that any common costs that may arise are
automatically allocated on the basis of resource consumption
45
The model
Cost drivers
The Oftel model uses a single increment for all traffic demand, representing a
single parallel increment for all traffic, plus an orthogonal increment for
customers

Busy hour total traffic load
Customers
Busy hour call attempts
Number of location updates
Peak SMS throughput
Coverage
Costs
Incremental
Traffic
Coverage
Services
Total cost of the network is taken to be the
sum of:

the standalone cost of providing a
specified level of coverage

the incremental cost of expanding that
network to carry a specified volume of
traffic

the incremental cost of expanding that
network to serve a specified volume
of customers
46
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
47
The model
Define what the
drivers of cost are
Define the
associated
services and increments
Cost
drivers,
services
and
increments
Define how the
increments
will interact
Services and increments
At one stage the Oftel model contained
eight separate services

In general, services should relate to the
fundamental services which the subscribers are
purchasing

Applications or value-added layered services are
not considered:

this simplification is influenced by the fact
that the vast majority of current network
traffic and costs are due to simple voice
communication

data transport is assumed to become more
important in later years, however we use a
Mbyte data transport service, rather than a
range of uncertain data applications
Handsets
Customers
Mobile originated off-net minutes
Mobile originated on-net minutes
Mobile terminated minutes
SMS messages
GPRS Mbytes
HSCSD Mbytes

The handset increment can be considered
separately from (i.e. is orthogonal to) the other
increments
48
The model
Services and increments
Considering all permutations of service demand requires a large number of
calculations (16 calculations for 4 increments)
Raw incremental costs
Voice
e.g. 80%
GPRS
HSCSD
SMS
1,2
3
Voice + SMS
Voice +…
4
5
6
Voice +…
7
8
9
10
11
GPRS + HSCSD
… + GPRS
.. + HSCSD
SMS + GPRS
SMS +…
.. + HSCSD
Voice + SMS + GPRS
Common costs
e.g. 5%
SMS + GPRS + HSCSD
Voice + SMS + …
.. + HSCSD
Voice +…
… + GPRS + HSCSD
Voice + SMS + GPRS + HSCSD
Coverage cost
Coverage
e.g. 15%
Areas are not to scale
Voice represents customers and voice minutes
Fully and separately resolving 8 increments would require 64 separate calculations
Incremental
costs using
single traffic
increment
49
The model
Services and increments
Even when the permutations have been calculated, the mark-up regime
becomes horrendous

Each common cost 1–11 needs to be
marked-up across the services which
it supports

The order and nature in which costs
are marked-up must be defined:
Voice
GPRS
HSCSD
SMS
1,2
3
Voice + SMS
Voice +…
4
5
6
Voice +…
7
8
9
10
11
GPRS + HSCSD
… + GPRS

for example, equalproportionate?

mark-up on mark-up?
.. + HSCSD
SMS + GPRS
SMS +…
.. + HSCSD

The sum of all the common costs 1–
11 is small in comparison with the
raw incremental costs of the major
traffic increments (voice and latterly
GPRS)

The coverage cost (by far the largest
common cost) must also be marked
up in some fashion
Voice + SMS + GPRS
SMS + GPRS + HSCSD
Voice + SMS + …
.. + HSCSD
Voice +…
… + GPRS + HSCSD
Voice + SMS + GPRS + HSCSD
Coverage
50
The model
Define what the
drivers of cost are
Define the
associated
services and increments
Cost
drivers,
services
and
increments
Services and increments
Hence, after investigation, we implemented
a single increment for traffic in the Oftel model
Define how the
increments
will interact


The model calculates incremental costs for
the services using a single increment

This increment resolves the allocation of
costs using routeing factors:

shared infrastructure on the basis of
demand consumption:
– equivalent voice equivalent
erlangs, or other parameter

dedicated infrastructure is still
allocated directly to the appropriate
service

This model enables:

understanding of the relevant increment
calculations

comparatively rapid calculation time

simple (yet automatic) allocation of
common costs between services

simplified mark-up step
And produces results for the voice LRICs that
are very close (~1% difference) to those of a
combinatorial multi-increment model
51
The model
Services and increments
In addition, the definition of the coverage network was altered …

The coverage network is required to:


Such a network, due to equipment divisibility, actually contains enough capacity to support
many more voice calls at no additional cost


support at least one incoming or outgoing voice call, anywhere within the coverage area
of the network
for example, one TRX has 8 channels
The coverage network was investigated. It was determined that:

a large proportion of the cost of the coverage network was actually equipment which
directly supported traffic or customers

only some equipment represented an absolute minimum requirement to provide coverage
– for example, the acquisition and preparation of the 2000–3000 sites required to
achieve minimum population coverage
52
The model
Services and increments
… to better reflect the relationship between capacity and cost
Coverage network
Macro-cell
site and TRXs
HLR
Backhaul
transmission
BTS
BSC–MSC
transmission

Inter-switch
transmission
The coverage network was broken into
two parts:

MSC
VLR
BSC
– network management
system (NMS) and points
of presence (macro site
acquisition, preparation
and rental)
NMS
Minimum coverage presence
Macro-cell
site acquisition,
preparation and rental

NMS
HLR
Backhaul
transmission
BTS
BSC–MSC
transmission
BSC
the coverage capacity
– equipment deployed in the
coverage network
providing more capacity
than actually required to
support just one voice
minute
Coverage capacity
Macro-cell
BTS and TRXs
the minimum coverage presence
Inter-switch
transmission
MSC
VLR
53
The model
Services and increments
The two parts of the coverage network are dealt with separately

The minimum coverage presence is used as the mark-up term

The coverage capacity is added to the incremental network capacity:


all capacity-providing elements deployed in the coverage network are considered as
incremental to traffic or customers as appropriate

the cost of these capacity elements is allocated according to routeing factors
This definition reduces the amount of cost in the coverage network, and as a consequence,
reduces importance of the choice of mark-up mechanism
54
The model
Services and increments
The Oftel model is a good representation of reality and significantly more
manageable than possible alternatives
Combinatorial multiple increment
Single increment, MCP
Voice
GPRS
HSCSD
Voice
SMS
GPRS
Voice + SMS
Voice +…
Voice +…
GPRS + HSCSD
… + GPRS
.. + HSCSD
SMS + GPRS
SMS +…
.. + HSCSD
Voice + SMS + GPRS
SMS + GPRS + HSCSD
Voice + SMS + …
.. + HSCSD
Voice +…
… + GPRS + HSCSD
Voice + SMS + GPRS + HSCSD
HSCSD
SMS
Coverage
Minimum coverage presence
Voice represents customers, incoming minutes, outgoing off-net and outgoing on-net minutes
Diagrams not to scale. Total cost is the same in both cases
55
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
56
The model
Demand forecasts
Demand forecasts are required in order to calculate cost results to 2006

It is important that this forecast is consistent with the methodology used elsewhere in the
model for determining the LRICs:


We primarily require a set of reasonable forecasts which will enable the model to be run,
investigated and produce reasonable information:


for example, the allowance for reasonable growth which is factored into the LRIC
approach should be consistent with the demand growth assumed in the forecasts
the assumption set was tailored to provide the required fidelity in forecasting, yet small
enough to be easy to use and modify
The forecasts used in the model were intended to be operator non-biased, for example:

all operators tend to the same market share

all operators are subject to the same rates of long term traffic growth

all operators have identical assumptions concerning HSCSD, GPRS and UMTS demand

historic nature of an operator’s subscriber base persists in the forecast
The model
57
Base case demand forecast: subscribers
Demand forecasts
The model
58
Demand forecasts
Base case demand forecast: outgoing minutes per subscriber per quarter
The model
59
Demand forecasts
Base case demand forecast: incoming minutes per subscriber per quarter
60
The model
Demand forecasts
The forecasts contain a number of inputs, calculations and outputs
Inputs take the form of:

S-curves, for parameters which grow to a
saturation point

Simple percentages for time dependent shares or
divisions

Quarterly growth rates, for parameters which
increase or decrease in a smooth fashion

The following demand parameters are calculated:

Mobile subscribers, by operator

SMS messages

Incoming and outgoing* voice minutes, on 2G
and 3G networks

HSCSD, GPRS and UMTS transport service
users and Mbytes of traffic

Demand parameters in future years, in order to
calculate allowances for reasonable growth
Outputs of the forecast are:

Demand parameters in each year
*outgoing voice minutes forecast includes outgoing on-net minutes
61
The model
Demand forecasts
S-curves are used for parameters which grow to a saturation point

x(t)
The inputs required for an s-curve are:

saturation of x

base year

x(A) at time A

x(B) at time B
saturation of x

x(B)
x(A)
t
base
A
B
Used for:

mobile market penetration

migration of voice traffic from 2G to 3G

data transport service penetration
62
The model
Demand forecasts
Percentage inputs are used for time dependent shares or divisions
100%

A simple percentage is used to distribute a
parameter across different categories

Used for:
80%

market shares

Mbytes across GPRS, HSCSD and UMTS
70%
60%
50%
40%
30%
20%
10%
UMTS
GPRS
HSCSD
Jul-09
Jul-08
Jul-07
Jul-06
Jul-05
Jul-04
Jul-03
Jul-02
Jul-01
0%
Jul-00
Partition of Mbytes by technology
90%
63
The model
Demand forecasts
Quarterly growth rates are used for parameters which increase or decrease in a
smooth fashion
growth 3
growth 2
growth 1
x(t)

Simple exponential growth (or decline) can be
specified with a single percentage

Annual growth rates are the compound of
quarterly growths:


t0
t1
t2
t3
t
e.g. 2% per quarter constitutes 8.2%
annually
The input of quarterly growth rates are used to
forecast:

minutes per subscriber

SMS per user

Mbytes per user
64
The model
Demand forecasts
Various levels of dimensionality are contained in the Oftel model forecasts
Mobile subscribers
by year quarters
by operator
Incoming voice
minutes
by year quarters
by operator
by technology
2G/3G
Outgoing voice
minutes
by year quarters
by operator
by technology
2G/3G
SMS messages
by year quarters
by operator
Data transport users
by year quarters
identical for
each operator
by technology
HSCSD, GPRS, UMTS
Data transport Mbytes
by year quarters
identical for
each operator
by technology
HSCSD, GPRS, UMTS
65
The model
Demand forecasts
Our technology assumptions by their nature contain implicit consideration of
the range of issues that will affect traffic on these networks


For example, the partition of voice traffic across 2G and 3G networks implicitly makes
assumptions on:

numbers of subscribers on 2G or 3G plans

operator strategies for 3G voice and data

3G coverage extent or black-spots

high-use 3G early adopters and low-use price-sensitive 2G remaining subscribers
The use of quarterly assumptions assist in defining accurately when services are assumed to be
launched
66
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
67
The model
Network design
The Oftel model network design algorithms are based on a
number of principles

Reflect industry practice with regard to base station layout, checked against existing networks

this checking is a combination of parameter calibration and application of industry
experience

Represent the use of modern technology

Satisfy the requirements of coverage and demand

Allow for reasonable growth (but no more)

Contain the key differences between GSM 900 and GSM 1800 radio network deployment


different cell radii and different radio layer spectral efficiency*
The model contains around 60 different units of equipment, sufficient to capture the required
fidelity in network design, yet small enough to be manageable
* Spectral efficiencies vary between GSM 900 and GSM 1800 networks in the UK because the 900MHz spectrum allocation is more fragmented
than the 1800MHz spectrum allocation
68
The model
Network design
Simplified network diagram
Macro-cell
site and TRXs
HLR
Backhaul
transmission
BTS
Backhaul
transmission
BSC–MSC
transmission
Inter-switch
transmission
BSC
MSC
VLR
PCU
NMS
Micro-cell or
pico-cell site
SGSN
Dedicated GPRS
infrastructure
* For the purposes of inter-switch transmission, we
assume BSC, SGSN and MSC are co-located
GGSN
Internet
69
The model
Network design
We define a number of area and cell types
Four area types
Six cell types
Rural
Urban
Omni
macrocell
Microcell
Suburban
Bi-sectored
macrocell
Highway
Tri-sectored
macrocell
Picocell
Tri-sectored GSM
1800 dual spectrum
overlay
70
The model
Network design
The area types are based on population densities

It is assumed that population density is a proxy to radio planning area types. Hence the model
utilises data from around 9000 postcode sectors to assist in the categorisation of area types in
this fashion

Definitions of the area types used are as follows:

Urban – postcode sectors with a population density larger than 8178

Suburban – postcode sectors with a population density between 8175 and 721 per km2

Rural – postcode sectors with a population density less than 721 per km2

Highway – 50% (11 000 km) of the primary roads in the UK
– this area type is actually “rural highways” since urban and suburban road are
assumed to be within urban or suburban coverage
71
The model
Network design
The six cell types allow differences in network design
by area type to be reflected

The model contains a number of inputs for:


the proportion of cells of each macro type, in each area type
However, the inputs currently assigned in the model reflect a simplified situation:

tri-sectored macro sites are deployed in urban and suburban areas

bi-sectored macro sites are deployed in highway areas

omni-sectored macro sites are deployed in rural areas

These cell types are deployed in response to the greater of coverage or traffic requirements

Micro and pico sites (defined as single sector, 2 and 1 TRX respectively) are deployed in response only to
traffic requirements, and furthermore, only in urban and suburban areas:


the amount of traffic that is carried on these cell layers is specified by a percentage (by area type) of
total traffic in that area type
In the UK, the GSM 900 operators also have GSM 1800 spectrum.. The model deploys a GSM 1800 layer
upgrade to these operators’ urban macro sites, and expands this in response to the amount of traffic loaded
onto this cell layer
72
The model
Network design
The use of a single increment for traffic requires
service demand drivers to be added together

In reality, the radio interface responds differently to voice circuit, GPRS packet and signalling
traffic. However, constructing a complex radio engineering model which separately deals with
these traffic types is not recommended in a LRIC costing exercise

Rules are required to combine traffic from voice, SMS, HSCSD and GPRS in a suitable way

The Oftel model contains voice equivalent erlangs:

the amount of traffic equivalent to one voice erlang

rules are defined for converting service demand (eg SMS messages, GPRS Mbytes) into
voice equivalent erlangs

Voice equivalent erlangs can then be added to normal voice erlangs, in order to drive the
network design algorithms with the aggregate traffic load

It is assumed that all services have a coincident busy hour (which may lead to some
overstatement of costs), and highly complex effects (such as different link margins (i.e. cell
radii) for GPRS traffic) are neglected
73
The model
Network design
SMS and HSCSD voice equivalent erlangs (VEErl)
SMS messages
40 bytes per SMS
HSCSD mbytes
voice channel rate of
767 bit/s
80% of user demand
in the downlink
1 minute  82 SMS
70% channel
occupancy


SMS messages are carried by signalling channels
in the radio layer:


the model assumes an average size for each
SMS message, and a data rate for a channel
the model assumes the use of SDCCH
(synchronous data control channel) for
SMS message transfer

voice channel rate of
14.4 kbit/s
1 minute  0.135
HSCSD
Mbyte
User demand represents both up and downlink
traffic
HSCSD is a circuit switched dialup data service that
enables users to open more than one channel in a
particular direction, in order to obtain a higher rate of data
transfer:

it is very similar to a circuit switched voice service

we need to assume a channel occupancy and data
rate
74
The model
Network design
GPRS voice equivalent erlangs

80% of user demand
in the downlink
100% channel
occupancy
voice channel rate of
9.05kbit/s (CS1)
12% additional
IP overheads
an allowance for
packetised nature
1 minute  0.09
GPRS Mbyte

User demand represents both up and downlink traffic

GPRS is an IP packet switched service:

hence the model assumes 100% channel
occupancy, and 12% overheads for IP protocol

GPRS has variable data rates:

four data rates (CS1–CS4) are available
with GPRS

CS4 (around 22kbit/s per channel)
represents transmission under idealised
conditions, or when the network has a low
level of loading

CS1 represents the lowest data rate of
transmission, and is the likely rate achieved
in the network under busy conditions (from
which the model is driven)
An allowance has been made for the ability of the
packetised GPRS service to utilise some of the
gaps in traffic which occur as a direct result of
using the erlang transformation to provision more
channels than required. This assumed allowance
is calculated to be small
75
The model
Network design
GPRS traffic can utilise (to an extent) gaps between voice conversations

A certain amount of ‘under-utilised’ capacity exists as a result of applying the erlang blocking
probability formula to the voice calls in a sector:

a BTCellnet paper (obtained from its website) indicates that this spare capacity can in fact
be used by GPRS:
– some probability should be applied to this spare capacity, to work out its effective
erlang capacity

The model calculates the difference between the number of channels deployed and the number
of erlangs supported:

this number of channels is used to determine the relative loading of voice circuits and
GPRS packet traffic:
– this factor is calculated to be 95%. i.e. GPRS packet traffic only demands 95% of
the capacity for the same amount of voice circuit switched traffic
The model
76
Network design
GPRS service demand interacts with a number of dedicated and traditional
GSM network infrastructure
SGSN100
Total GPRS BH kbit/s
(+12% IP)
GPRS subscribers
GGSN100
IP transmission
PCU
Downstream GPRS BH kbit/s
Dedicated GPRS
infrastructure
Downstream GPRS voice
equivalent BHE
Existing GSM
infrastructure
Air interface
Backhaul
BH = busy hour
77
The model
Network design
The HSCSD service places demands upon all traditional GSM infrastructure
Radio and transmission

HSCSD voice equivalent erlangs are added to
voice circuit switched erlangs (using routeing
weighting) and used to drive the deployment of
traditional GSM infrastructure, including:

base station sites and TRX

backhaul

BSC switching

interswitch transmission

switch ports
MSC/VLR processing

Voice calls require MSC/VLR processing to
originate and terminate. This processing includes
checking the validity of the subscriber, and
locating the mobile handset in the network

HSCSD calls also require processing when they
are originated from a HSCSD enabled handset:

we assume an average HSCSD session of
0.25Mbyte

assume 1.1 session attempts per session

assume the same MSC/VLR processing per
session as an outgoing voice call attempt
(20ms)
78
The model
Network design
Additional allowance for the distribution of traffic is made,
over and above the use of area types

The model currently contains four area types (urban, suburban, rural, highway) in order to
distribute traffic load across the country in a sensible fashion

However, within each area type, demand will be distributed non-homogeneously (both in time
and space), and an allowance for this is included

The requirement for half an additional unit of capacity at each point in the network was
calculated by Analysys using a network simulation tool
Area type
Diagrams not to scale
Area type
Highway
Suburban
Urban
Non-homogeneous reality
erlangs per sector
Highway
Rural
Suburban
Urban
erlangs per sector
Simplified average situation
Rural

to account for this effect, an additional ½ TRX is deployed on each sector
Additional capacity
requirement over
the average
79
The model
Network design
Equipment utilisation is an important input parameter
to the network design algorithms

A large number of network design calculations are based upon the following relationship:


number of items required = demand / capacity per item * utilisation
The utilisation parameter contained in the Oftel model is used to reflect the explicit combination of a number
of different ‘under-utilisation’ effects:

Design utilisation: most equipment has a (vendor designated) maximum utilisation parameter (for
example, 90%). This utilisation parameter ensures that the equipment in the network is not overloaded
by any transient spikes in demand

Scorched node utilisation: the deployment of a scorched node network is captured explicitly by the
use of additional utilisation parameters. These indicate the degree to which equipment is unable to
reach the level of utilisation that would be achieved in a scorched earth deployment, as a direct result
of adhering to the scorched node constraint

Reasonable growth utilisation: in a real mobile network, equipment is deployed in advance of
expected demand (weeks to years), depending on the equipment modularity and the time it takes to
make all the necessary preparations to bring new equipment online. The model explicitly determines
the level of under-utilisation in the network, as a function of equipment planning periods and expected
demand.
80
The model
Network design
Reasonable growth utilisation parameters are calculated explicitly

Explicit inputs relating to the provision of a reasonable allowance for future growth enable the
effect on average equipment utilisation to be calculated

This is done for a number of asset classes, by choosing:

the key demand driver which is to be used in determining future growth in demand

the point in the future at which demand should be considered
– The future demand point for each asset class is taken to be half of one planning
period in the future, based on the simple assumption that some sites will have only
just been upgraded (and hence have sufficient capacity to meet demand anticipated
one entire planning period into the future) whereas other sites will be about to be
upgraded (and therefore are only able to meet current demand), with most sites lying
somewhere in between these two extremes (and hence on average the effect is likely
to be as if all sites have sufficient capacity to meet demand for about half of one
planning period into the future)

The model contains a forecast of demand over time, which is then used in the calculation of the
reasonable growth utilisation
81
The model
Network design
Calculation method for reasonable growth utilisation
Assign a key driver to each class of infrastructure, e.g. demand:


define planning period (2p), and determine demand at time half planning
period later:


= xt / (capacity * normal utilisation)
Number of elements deployed at time t+p, if no future growth:


(demand at time t + p ) = xt+p
Number of elements deployed at time t, if no future growth:


(demand at time t) = xt
Demand

= xt+p / (capacity * normal utilisation)
Hence actual utilisation of elements at time t, given forward looking
deployment is:

xt+p / (capacity * normal utilisation) = xt / (capacity * actual utilisation)

hence:
– actual utilisation = normal utilisation * (xt / xt+p)
 normal utilisation = design utilisation * scorched node utilisation allowance
t
t+p
time
82
The model
Network design
An example of maximum utilisation


Macrocell BTS:

design utilisation
input at 80%

scorched node allowance
input at 90%
Due to the inefficiencies which
arise as a result of scorched
node (compared to scorched
earth) BTSs are not able to reach
their designed utilisation

Reasonable growth driver
set to “traffic”
The main driver of the
deployment of BTSs is traffic

Look-ahead
selected as 2 years ahead
Traffic in two years time is 60% higher than today’s traffic, hence


Vendor says “do not run a BTS at
more than 80% peak capacity”
reasonable growth allowance = 1/1.6 = 63%
Calculated maximum utilisation of a macrocell BTS is thus:

80% * 90% * 63%

= 45%
BTSs (sites) have a long
planning period
83
The model
Network design
For each asset class, the key demand driver and
period of planning must be selected
Asset classes

TRX

BTS – macro, micro and pico

backhaul links
Key demand drivers
Look-ahead period

Year average subs

Current time

Year total incoming minutes

2 weeks ahead

Year total outgoing minutes

1 month ahead

Year total SMS messages

1 quarter ahead

Year total GPRS Mbytes

6 months ahead

Year total HSCSD Mbytes

1 year ahead

BSC

BSC-MSC transmission

MSC/VLR – CPU and ports

HLR

Inter-switch transmission

Year average GPRS users

2 years ahead

SMSCs

Year total minutes

3 or more years ahead

PCU

Year total approx traffic


GSNs – connections and peak
throughput
IP transmission
84
The model
Network design
The model also explicitly calculates the output utilisation profiles required for
the economic depreciation calculations

The economic depreciation calculations require equipment utilisation profiles (taken into
account when calculating economic life and distributing the cost of an asset over its lifetime)

These profiles are calculated for a number of classes of equipment in the model

The reasonable growth utilisation factor is not taken into account in the determination of
output utilisation since these assets are deployed in advance of the demand they will support
100% utilisation
Design utilisation allowance
Scorched node utilisation allowance
100%
y
Actual out-turn utilisation
x(t)
Output utilisation
profile for
economic
depreciation is
x(t) / y
time
85
The model
Network design flow diagrams

The following slides provide details of the network design algorithms:

flow diagrams

explanatory sections relating to these flow diagrams
Input parameter
(data or assumption)
Calculation
Major equipment deployment
output
Network design
86
The model
Network design
Base station sites
Maximum cell
radii
Spectrum
Reuse
TRX bandwidth
Spectral capacity of a sector
Area to cover
Maximum cell
area
Non-uniform
allowance (0.5 TRX/sector)
BTS and TRX unit capacity
Maximum achievable capacity
of a sector
Utilisation of TRX and BTS
Effective capacity of a sector
TRX Traffic (BHE)
Sectors required for capacity
Site type proportions
Sites (by type) required for
capacity
Sites required for coverage
Number of sites (by type)
used in TRX calculations
87
The model
Network design
Base station sites (2)
Maximum cell
radii
Spectrum
Reuse
TRX bandwidth
Spectral capacity of a sector
Spectrum, reuse and
TRX bandwidth are Maximum cell
Area towell
cover
reasonably
area
defined parameters
Non-uniform
allowance (0.5 TRX/sector)
BTS and TRX unit capacity
Maximum achievable capacity
of a sector
Utilisation of TRX and BTS
Effective capacity of a sector
The non-uniform allowance is the ½ unit
TRX
Traffic (BHE)
of capacity
per sector allowanceSectors
for the required for capacity
fact that traffic is not evenly distributed (in
both time and space) across each area
type
Sites (by type) required for
Site type proportions
capacity
Sites required for coverage
Number of sites (by type)
used in TRX calculations
88
The model
Network design
Base station sites (3)
Spectrum
Different cell radii are used
for each area type, and for
GSM 900 and GSM 1800.
Reuse
TRX traffic is the
(routeing weighted*) sum of
all the traffic types, allocated to each area
The area to cover is again
TRX
bandwidth
Spectral
capacity
a sector
and cell type using percentage inputs
by area
type,ofand
in terms of
2
km
Site type proportions are simplified
Maximum cell
radii
Area to cover
Maximum cell
area
Non-uniform
assumptions
for:
allowance (0.5 TRX/sector)
• all urban and suburban as tri-sectored
• all BTS
highway
bi-sectored
and as
TRX
unit capacity
• all rural as omni-sectored
Maximum achievable capacity
of a sector
• micro and pico sites are defined as
omni-sectored
Utilisation of TRX and BTS
Effective capacity of a sector
TRX Traffic (BHE)
Sectors required for capacity
Site type proportions
Sites (by type) required for
capacity
Sites required for coverage
Number of sites (by type)
* routeing weighted: for example, one on-net mobile-to-mobile minute has two
used in TRX calculations
contributions to TRX BHE
89
The model
Network design
Base station sites (4)
Maximum cell
radii
Spectrum
Reuse
TRX bandwidth
Spectral capacity of a sector
Area to cover
Maximum cell
area
Non-uniform
allowance (0.5 TRX/sector)
BTS and TRX unit capacity
Utilisation of TRX and BTS
TRX Traffic (BHE)
Site type proportions
Maximum achievable capacity
of a sector
The number of sites
Effective
a sector
deployedcapacity
(for eachofarea
and
cell type) is determined as
the greater of those required
for coverage or traffic
Sectors required for capacity
Sites (by type) required for
capacity
Sites required for coverage
Number of sites (by type)
used in TRX calculations
The model
90
Typical results of Base station site calculations
Network design
91
The model
Network design
TRXs
Number of sites
Sectors per site (by site type)
Number of sectors
TRX traffic (BHE)
Traffic per sector (BHE)
TRX unit capacity and
utilisation
TRXs per sector to meet traffic
requirements
from Sites calculations
Non-uniform
allowance (0.5 TRX per sector)
Minimum TRXs per sector
Number of TRXs per sector
Number of TRXs (all sectors)
used in Site–BSC transmission calculations
used in BSC calculations
92
The model
Network design
TRXs (2)
Number of sites
Sectors per site (by site type)
TRX traffic (BHE)
TRX unit capacity and
utilisation
from Sites calculations
Number of sectors
These
assumptions
areTraffic
the same
per sector (BHE)
as used in the
BTS
calculations
TRXs
per sector to meet traffic
requirements
Non-uniform
allowance (0.5 TRX per sector)
Minimum TRXs per sector
The minimum
TRX
deployment
Number
of TRXs per sector
is 1 TRX per
sector
Number of TRXs (all sectors)
used in Site–BSC transmission
calculations
used in BSC calculations
93
The model
Network design
TRXs (3)
Number of sites
from Sites calculations
Sectors per site (by site type)
Number of sectors
TRX Traffic (BHE)
Traffic per sector (BHE)
TRX unit capacity and
utilisation
TRXs per sector to meet traffic
requirements
The final number of TRXs is
again calculated in response to
coverage requirements (driven
by the number of sites) and
traffic requirements (driven by
the amount of traffic per sector)
Number of TRXs per sector
Number of TRXs (all sectors)
used in Site–BSC transmission calculations
used in BSC calculations
Non-uniform
allowance (0.5 TRX per sector)
Minimum TRXs per sector
The model
94
Typical results of TRX calculations
Network design
95
The model
Network design
Base station site – BSC transmission
Number of TRXs per sector
Required circuits per TRX
Required circuits per sector
Sectors per site (by site type)
Required circuits per site
Link capacity
(by link rate)
Links required per site
(by link rate)
Link utilisation
Links required per site
(at selected link rate)
Link type proportions
Leased lines
(by link rate)
from TRX calculations
Microwave links
(by link rate)
used in BSC calculations
Hops per link
Microwave hops (by link rate)
96
The model
Network design
Base station site – BSC transmission (2)
Number of TRXs per sector
Required circuits per TRX
Required circuits per sector
Sectors per site (by site type)
Required circuits per site
Link capacity
(by link rate)
Links required per site (by link
rate)
Link utilisation
Links required per site
(at selected link rate)
Link type proportions
Leased lines
(by link rate)
from TRX calculations
Microwave links
(by link rate)
used in BSC calculations
The number of circuits per TRX
is a well known network design
parameter
A calculation determines the
number of links of each type (2,
8, 16, 32 Mbit/s) required to
support the demand …
… and then deploys no more
Hops
link
than one link
perper
site,
selecting
the required link capacity
Microwave hops (by link rate)
97
The model
Network design
Base station site – BSC transmission (3)
Number of TRXs per sector
Required circuits per TRX
Required circuits per sector
Sectors per site (by site type)
Required circuits per site
Link capacity
(by link rate)
For example, 80% microwave
self provided and 20% leased
lines, specified for macro,
Link utilisation
micro and pico sites in each
area type
Link type proportions
from TRX calculations
Links required per site (by link
rate)
Links required per site
(at selected link rate)
Leased lines
(by link rate)
Microwave links
(by link rate)
used in BSC calculations
Again, specified for macro,
micro and pico sites in each
area type
Hops per link
Microwave hops (by link rate)
98
The model
Network design
BSCs
from TRX calculations
Number of TRXs
(all sectors)
Number of BSCs
BSC capacity
Utilisation
from Site – BSC calculations
from BSC – MSC calculations
Leased lines
(by link rate)
Leased lines
(by link rate)
Microwave links
(by link rate)
Microwave links
(by link rate)
Number of BTS-facing ports
Number of MSC-facing ports
Ports per link (by link rate)
Ports per link (by link rate)
used in BSC – MSC
transmission calculations
used in MSC calculations
99
The model
Network design
BSCs (2)
from TRX calculations
Number of TRXs
(all sectors)
from Site–BSC calculations
Leased lines
(by link rate)
Microwave links
(by link rate)
from BSC – MSC calculations
Leased lines
(by link rate)
Microwave links
(by link rate)
BSC deployments are simply
driven by the number of TRXs
deployed in the radio network
Number of BSCs
BSC capacity
Utilisation
Number of BTS-facing ports
The number of BSC ports does
not drive the deployment of
BSCs, but the number of MSCfacing ports is taken into
account in the MSC
Ports
per link (by link rate)
dimensioning
used in BSC – MSC
transmission calculations
Number of MSC-facing ports
Ports per link (by link rate)
used in MSC calculations
100
The model
Network design
BSC – MSC transmission
Number of BSCs
BSC–MSC traffic (BHE)
Traffic per BSC
Link capacity
(by link rate)
Links required per BSC
(by link rate)
from BSC calculations
Link utilisation
Links required per BSC
(at selected link rate)
Link type proportions
Leased lines
(by link rate)
Microwave links
(by link rate)
used in BSC calculations
Hops per link
Microwave hops
101
The model
Network design
BSC – MSC transmission (2)
Number of BSCs
BSC – MSC traffic (BHE)
Traffic per BSC
BSC – MSC
is again a
Linktraffic
capacity
(routeing (by
weighted)
sum of all
link rate)
traffic types passing from BSC
to MSCs
Link utilisation
Links required per BSC
(by link rate)
from BSC calculations
Links required per BSC
(at selected link rate)
Link type proportions
Leased lines
(by link rate)
Microwave links
(by link rate)
used in BSC calculations
Hops per link
Microwave hops
102
The model
Network design
BSC – MSC transmission (3)
BSC–MSC traffic (BHE)
These calculations are similar
Number
ofbase
BSCsstation
to those
used for
site – BSC transmission,
though involve different
assumptions where appropriate
Traffic per BSC
Link capacity
(by link rate)
Links required per BSC
(by link rate)
from BSC calculations
Link utilisation
Links required per BSC
(at selected link rate)
Link type proportions
Leased lines
(by link rate)
Microwave links
(by link rate)
used in BSC calculations to define number of MSC-facing
ports required (but not the number of BSCs)
Hops per link
Microwave hops
103
The model
Network design
MSCs
from BSC calculations
Interconnect traffic (BHE)
Number of MSC-facing ports
Minimum interconnect ports
MSC capacity (ports)
Number of BSC-facing ports
Switch port capacity
Number of MSCs required to
meet demand for ports
Switch port utilisation
Minimum MSCs
Number of interconnect-facing
ports
CPU capacity (BHms)
Total number of ports
CPU utilisation
Interswitch traffic (BHE)
MSC capacity (CPUs)
Number of interswitch ports
Processing demand (BHms)
Switch port capacity
Switch port utilisation
Number of MSC/VLRs
used in MSC transmission calculations
104
The model
Network design
MSCs (2)
Interconnect traffic (BHE)
from BSC calculations Number of MSC-facing ports
Minimum interconnect ports
MSC capacity (ports)
Number of BSC-facing ports
Number of MSCs required to
meet demand for ports
Minimum MSCs
CPU capacity (BHms)
Switch port
capacity
Switch port utilisation
MSC/VLRs are deployed in response to the
Number
of interconnect-facing
CPU processing requirements of the
network,
ports
generated by a number of services and
processes, including:
number of ports
• subscriberTotal
authentication
CPU utilisation
MSC capacity (CPUs)
• incoming and outgoing circuit switched call
set-ups
• SMS message send and delivery
Number of interswitch ports
• subscriber location updating
Processing demand (BHms)
Interswitch traffic (BHE)
Switch port capacity
Switch port utilisation
Number of MSC/VLRs
used in MSC transmission calculations
105
The model
Network design
MSCs (3)
Interconnect traffic (BHE)
from BSC calculations Number of MSC-facing ports
Minimum interconnect ports
MSC capacity (ports)
Number of BSC-facing ports
Number of MSCs required to
meet demand for ports
Switch port
capacity
Switch port utilisation
Minimum MSCs
Number of interconnect-facing
ports
CPU capacity (BHms)
Total number of ports
CPU utilisation
Interswitch traffic (BHE)
In addition, the number of MSCs should
also have sufficient capacity to support
Number
of demands.
interswitch ports
port
MSC capacity (CPUs)
Switch port capacity
However, this link is not automatic in the
model, and must be completed with a
Switch port utilisation
manual check.
Processing demand (BHms)
Number of MSC/VLRs
used in MSC transmission calculations
106
The model
Network design
MSCs (4)
from BSC calculations
The number of ports are summed up
from the three major types of ports
MSC
present
incapacity
the MSC(ports)
Interconnect traffic (BHE)
Number of MSC-facing ports
Minimum interconnect ports
Number of BSC-facing ports
Switch port capacity
• each MSC-facing port in a BSC
Number
requires a reciprocal
port in of
theMSCs
MSC required to
meet demand for ports
• interconnect ports are driven by
interconnect traffic (routeing weighted
MSCs
sum of Minimum
all relevant
traffic types), capacity
and utilisation inputs
• there
may
be a contractual
CPU
capacity
(BHms) QoS
minimum requirement for the number of
interconnect ports
CPU utilisation
• interswitch ports are also driven by
interswitch traffic (routeing weighted sum
of allMSC
relevant
traffic
types), capacity and
capacity
(CPUs)
utilisation inputs
Switch port utilisation
Number of interconnect-facing
ports
Total number of ports
Interswitch traffic (BHE)
Number of interswitch ports
Processing demand (BHms)
Switch port capacity
Switch port utilisation
Number of MSC/VLRs
used in MSC transmission calculations
The model
107
Network design
Interswitch transmission
Number of interswitch ports
Transmission utilisation
Number of interswitch circuits
from MSC calculations
The model
108
Network design
Interswitch transmission (2)
Number of interswitch ports
Transmission utilisation
Number of interswitch circuits
from MSC calculations
The number of interswitch
ports is simply driven by the
number of interswitch ports
(which was in itself driven by
the amount of interswitch
traffic)
109
The model
Network design
HLR capacity
HLR capacity
Minimum number of HLRs
Number of HLRs
Number of customers
HLR utilisation
Number of HLR upgrades
HLR upgrade capacity
110
The model
Network design
HLR capacity (2)
HLR capacity
Minimum number of HLRs
Number of HLRs
Number of customers
HLRs are again driven by a
simple calculation involving
capacity, demand and
utilisation.
However, at least two HLRs
are required, at minimum,
for redundancy
HLR utilisation
Number of HLR upgrades
HLR upgrade capacity
111
The model
Network design
HLR capacity (3)
HLR capacity
Number of customers
Minimum number of HLRs
Capacity upgrades to the HLRs
are deployed, however the full
Number
costofofHLRs
a HLR is assumed in
the base HLR, and hence HLR
upgrades do not impact the
cost results
HLR utilisation
Number of HLR upgrades
HLR upgrade capacity
112
The model
Network design
Typical results of BSC, MSC and HLR calculations
140
120
100
80
BSCs
MSCs
HLRs
60
40
20
0
93/94
94/95
95/96
96/97
97/98
98/99
99/00
00/01
01/02
02/03
113
The model
Network design
SMS centres
Minimum number of SMSCs
SMS throughput demand
SMSC throughput capacity
Utilisation
Number of SMSCs
114
The model
Network design
SMS centres (2)
Minimum number of SMSCs
SMS throughput demand
SMSC throughput capacity
Utilisation
SMS throughput demand is
again a routeing weighted sum
of all SMS types:
• Mobile originated (MO) off-net
Number of SMSCs
• MO on-net
• MT
• server originated (voicemail,
info-service, etc)
115
The model
Network design
Dedicated GPRS equipment – PCU boards
from BSC calculations
Number of BSCs
GPRS MB throughput demand
PCU throughput capacity
Number of PCUs by throughput
Number of PCUs by 1 per BSC
minimum
Utilisation
Number of PCUs
116
The model
Network design
PCU boards (2)
GPRS MB throughput demand
PCU throughput capacity
The number of PCUs deployed
(packet control unit upgrades to
from BSC calculations
BSCs) is calculated as the
greater of capacity demands
or
Number
of BSCs
one per BSC
Number of PCUs by throughput
Number of PCUs by 1 per BSC
minimum
Utilisation
Number of PCUs
117
The model
Network design
Dedicated GPRS equipment – GGSNs
GPRS MB throughput demand
Active
GPRS PDP contexts
GGSN throughput capacity
GGSN
PDP context capacity
Throughput utilisation
PDP context utilisation
Number of GGSNs by
throughput
Minimum number of GGSNs
Number of GGSNs by PDP
contexts
Number of GGSNs
118
The model
Network design
GGSNs (2)
GPRS MB throughput demand
GGSN throughput capacity
The greatest of three requirements are
taken into account when calculating the
number of GGSNs deployed:
• at least two for redundancy
Active
GPRS PDP contexts
GGSN
PDP context capacity
• throughput traffic requirements
Throughput utilisation
• PDP context (IP address) requirements
Number of GGSNs by
throughput
Minimum number of GGSNs
PDP context utilisation
Number of GGSNs by PDP
contexts
Number of GGSNs
119
The model
Network design
Dedicated GPRS equipment – SGSNs
GPRS MB throughput demand
Connected GPRS subscribers
SGSN throughput capacity
SGSN
subscriber capacity
Throughput utilisation
Subscriber utilisation
Number of SGSNs by
throughput
Minimum number of SGSNs
Number of SGSNs by
subscribers
Number of SGSNs
120
The model
Network design
SGSNs (2)
GPRS MB throughput demand
SGSN throughput capacity
The greatest of three requirements are
also taken into account when calculating
the number of SGSNs deployed:
• at least two for redundancy
Connected GPRS subscribers
SGSN
subscriber capacity
• throughput traffic requirements
Throughput utilisation
• GPRS subscriber requirements
Number of SGSNs by
throughput
Minimum number of SGSNs
Subscriber utilisation
Number of SGSNs by
subscribers
Number of SGSNs
121
The model
Network design
Dedicated GPRS equipment – IP transmission
GPRS IP Mbit/s
Number of IP transmission
2Mbit/s links
Transmission utilisation
122
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
123
The model
Economic Depreciation
The problem

How would an operator set its prices if it were operating in a (hypothetical) fully competitive
and partially contestable market?

So as to neither under- nor over-recover costs, since:
– they would not enter if costs could not be fully recovered
– they would be prevented from over-recovery of costs by competition

Consistent with changes in the underlying costs of production and the contestability of
the market, since:
– they will set their prices in line with those that a new entrant into the market at each
point in time would charge

Traditional depreciation methods, such as straight-line or reducing balance depreciation, can
achieve the first of these requirements, but not in general the second.

Economic depreciation can achieve both.
124
The model
Economic Depreciation
Competitiveness vs Contestability

Competitiveness describes the extent to which operators already in the market compete with
each other (and thereby control each others behaviour):


A fully competitive market is one in which there are at least two (non-collusive) players
and no customer switching costs – customers can (and will) instantaneously switch from
one provider to another if a better deal is on offer
Contestability describes the ease with which operators can enter (and exit) the market (and
thereby control the behaviour of those already in the market):


A fully contestable market has no barriers to entry and exit – a new entrant can enter the
market and capture all of an incumbent’s existing customers instantaneously if they offer
a better deal
A partially contestable market has barriers to entry and exit – new entrants into the
market can only capture customers from the incumbent after some delay (for example the
time necessary to roll out their network) and/or at some limited rate (for example because
of the need to build up their reputation and brand image)
125
The model
Economic Depreciation
What difference does it make whether a market is fully
or only partially contestable?


In a fully contestable market, incumbents (players already in the market, irrespective of the
date they entered, or their scale) can never set prices higher than what it would cost a new
entrant to provide the same service, using the most efficient means, since:

If they were to do so, new players would enter, set lower prices, and capture the entire
market

(NB This is true even if the incumbent is a monopoly!)
In a less than fully contestable market, incumbents may be able to temporarily sustain prices
that are higher than what it would cost a new entrant to provide the same service, using the
most efficient means, to the same number of customers as the incumbent, since:


It will take time for the new entrant to be ready to capture all of the incumbents’
customers
And so in the mean time the new entrant’s cost per customer will be higher than it would
be if they had instantaneously captured the entire market
126
The model
Economic Depreciation
Why then can’t incumbents in a less than fully contestable market
over-recover their costs?

They can if the market is less than fully competitive!

But if the market is fully competitive (or assumed to be), competition between the incumbents
will ensure that prices overall (over the lifetime of the product) are no higher than the costs of
production, since if any one incumbent attempted to set a price, at any time, that was higher
than the competitive level, they would instantaneously lose all of their customers to their
competitors.

In a fully competitive market it is therefore only the timing of the recovery of costs that differs
between scenarios of full and partial contestability, not the total amount of cost recovered:

If the market is fully contestable, operators have to recover costs in each year from the
customers making use of the service in that year, which in theory would lead to very high
prices in the early years of operation

If the market is less than fully contestable then operators can keep prices at a reasonable
level in the early years, albeit with a compensatory but small increase in prices in later
years
127
The model
Economic Depreciation
Why model a less than fully contestable market?

Mobile markets are in practice less than fully contestable:

Significant up-front investment in network roll-out is required before any customers can
be signed up

It took time for mobile operators to build up the market for mobile services

If the mobile operators had set prices commensurate with a fully contestable market in the
early years those prices would have been very high, in which case the market would probably
have never developed

If mobile operators are now forced to set prices as if the market were fully contestable then
they will never fully recover the costs of their initial investments (they will suffer a so-called
“windfall loss”)
128
The model
Economic Depreciation
The economic depreciation problem restated

What time-series of prices, consistent with trends in the underlying costs of production and the
assumed contestability of the market, yield an expected NPV of zero over the period of
interest?

An NPV of zero ensures that the prices are cost-based, as they would have to be in a fully
competitive market, neither under- nor over-recovering total costs over the lifetime of the
project

Consistency of prices with trends in the underlying costs of production and assumed
contestability of the market ensure that those prices are reflective of those that a
(hypothetical) new entrant into the market at each point in time would charge
129
The model
Economic Depreciation
The inputs
Expenditure
1000
100%
800
80%
600
60%
400
40%
Output (utilisation)
Underlying cost trend
(capital and opex combined)
Capital investment
200
20%
Operating expenses
0
0%
0
1
2
3
4
5
6
7
Year of life
8
9 10 11 12
130
The model
Economic Depreciation
First calculate the total expenditure…
Expenditure
(We will initially assume a lifetime of 10 years)
1800
1800
1600
1600
1400
1400
1200
1200
1000
1000
800
800
600
600
400
400
200
200
0
0
0
1
2
3
4
5
6
7
Year of life
8
9
10 11 12
Capital investment
Operating expenses
PV of total expenditure
(up to year 10)
131
The model
Economic Depreciation
…then calculate the total relative output value
(assuming the same lifetime of 10 years)
350%
350%
300%
300%
250%
250%
Output (utilisation)
200%
200%
Underlying cost trend
(capital and opex combined)
150%
150%
Relative output value
100%
100%
PV of total relative output
value (up to year 10)
50%
50%
0%
0%
0
1
2
3
4
5
6
7
Year of life
8
9 10 11 12
132
The model
Economic Depreciation
Divide one by the other to yield the unit price
for a relative output value of 100%
334%
1644
492
PV of total expenditures
PV of total relative output value
Unit price at 100% output value
133
The model
Economic Depreciation
Multiply this by the relative output value in each year to yield annual revenues
500
90%
450
80%
400
70%
350
60%
300
50%
Unit price at 100%
output value
Relative output value
250
40%
200
Revenue
30%
150
100
20%
50
10%
0
0%
0
1
2
3
4
5
6
7
Year of life
8
9
10
11
12
134
The model
Economic Depreciation
Economic depreciation is then the difference between revenues and operating
expenses
Economic lifetime = last
year in which economic
depreciation is positive
500
400
Check that this matches
with earlier assumption
300
200
Revenue
100
Economic depreciation
Operating expenses
0
-100
-200
0
1
2
3
4
5
6
7
Year of life
8
9
10
11
12
135
The model
Economic Depreciation
Check that everything is consistent!
1,800
1,600
1,400
1,200
1,000
800
600
400
200
0
PV of total
revenues
Revenues
PV of total
annualised costs
Operating expenses
Economic depreciation
PV of total
expenditures
Capital investment
136
The model
Economic Depreciation
Notes re implementation in the
LRIC Model of UK Mobile Network Costs [1]


Model considers a period of interest longer than one asset lifetime:

Includes investment necessary to replace assets at the end of their useful life

Uses perpetuities to model the period beyond the finite horizon of the explicit
calculations
This is economically rational in a less than fully contestable market since operators invest for
the long term, not merely to obtain customers for the lifetime of each individual asset
137
The model
Economic Depreciation
Notes re implementation in the
LRIC Model of UK Mobile Network Costs [2]

Model calculates “revenue required” separately for capital investment and operating expenses


Simplifies modelling of separate underlying cost trends for capital costs (MEA prices)
and operating expenses
Operating expenses are assumed to vary both with time and age of asset
138
The model
Economic Depreciation
Notes re implementation in the
LRIC Model of UK Mobile Network Costs [3]


Model computes “revenue required” separately for each of three components of total cost:

Long-run equilibrium costs – based on long-run equilibrium input prices and output

Additional costs of lower output in earlier years

Additional costs of higher input prices in earlier years
Makes it easier to ensure that the “underlying cost trend” is consistent with the assumed
evolution of the market
139
The model
Economic Depreciation
Notes re implementation in the
LRIC Model of UK Mobile Network Costs [4]

The “underlying cost trend” applied in each case is different reflecting the different forces at
work in each case:

Long-run costs = long-run input cost trend

Costs of lower output = Extent to which later entrants achieve long-run output more
quickly than do earlier ones

Costs of higher input prices = Extent to which earlier entrants have to pay input prices
higher than those implied by the long-run trend
140
The model
Economic Depreciation
Notes re implementation in the
LRIC Model of UK Mobile Network Costs [5]

Model tracks history of UK operators to date, together with a forecast of their likely future
development

Would be equally valid to model the future of a new entrant into the market today (or any other
date), but:


This would entirely disconnect the model from the reality of the incumbent operators
(who are the ones whose charges are to be regulated)

Makes the model and results entirely dependant upon forecasts
Results ought to be the same anyway, since the objective of the approach is to identify that set
of prices which an incumbent would charge which are consistent with those that new entrants
would charge
141
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
The model
142
Network costing
Network costing is the multiplication of economic cost per item and network
deployment per item
Economic cost for each item
Coverage network deployment
Coverage
network
cost
+
Incremental network deployment
Incremental
network
cost
=
Full network deployment
Full network
cost
143
The model
Network costing
A number of business activities are included either directly or indirectly in the
network costs


Included explicitly as direct costs:

equipment, site rentals, switch software, building preparation

network management
Included as indirect costs, per unit of infrastructure:

maintenance

accommodation, power, vehicles and IT
144
The model
Network costing
A number of similar calculations produce coverage, incremental
and total costs
Economic
cost per
item
x
Total
Incremental
Total
Incremental
MCP
MCP
Number
of items
deployed
Total cost
for each
item
=
1
1
1
2
2
2
3
3
3
…
…
…
Total cost

The simple multiplication of economic cost per
item and equipment deployments produces the
headline total costs:

coverage network cost, defined as just the
MCP

incremental network cost (which includes
the equipment designated as coverage
capacity)

total network cost
145
The average incremental cost per unit output of each network element is
simply the incremental cost of each network element divided by its output
Demand by service
1 2 3 …
x
Incrementalc
ost of each
network
element
1

Routeing factors
=
Average
incrementalc
ost per unit
output of
each
network
element
2
1
3…
2
3…
=
Output of each
network element
146
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
147
The model
Service costing
The matrix of routeing factors is key to the allocation of incremental costs to
services
Routeing factors*
Average incremental cost per unit
output of each network element
Unitised
incremental
cost per
service
Common costs of coverage
Mark-ups
to recover
common
costs
* Routeing factors defined below
148
The model
Service costing
Routeing factors are relative numerical weightings for the consumption of
resources by services
Routeing factors
=
Capacity of each
network element
required by each
service (per
unit of demand)
149
The model
Service costing
The axes of the routeing factor matrix are services and network elements
OG on-net mins
SMS messages
1
1
2
0.01
BSC
1
1
2
0.01
HLR
MSC/VLR
1
1000
etc
OG off-net mins
3-sector macro
Assets
Customers
Incoming mins
Services

Each network cost must be allocated to one or
more services, according to the consumption of
resources

The allocation of certain network costs to
particular increments may be varied, provided
there is a good reason for allocating such a cost to
a different service increment than that used to
drive the cost. For example:

50
20
70
etc
* Illustrative routeing factors; OG = outgoing
the costs of location updates could be
allocated to customers or traffic, depending
on whether location updates were seen as a
feature applicable to subscribers or calls
150
The model
Service costing
The output of the service costing calculation is unitised incremental service
costs
Average
incrementalc
ost per unit
output of
each
network
element
x
Routeing factors
=
Unitised
incremental
cost per
item, for
service 1
1
1
2
2
3…
3…
Total
unitised
incremental
cost for
service 1
Unitised
Unitised
Unitised
incremental
Unitised
incremental
Unitised
incremental
cost
per
incremental
cost
per
incremental
cost
per
item,
for
cost
per
item,
for
cost
per
item,
for
service
2
item,
for
service
2 2for
item,
service
service
22
service
11
11
1
22
22
2
3…
3…
3…
3…
3…
…
…
Unitised
incremental
cost per
item, for
service 8
1
2
3…
Total
unitised
incremental
cost for
service 8
151
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
152
The model
Results
The key outputs of the model are unitised and total costs
For each year of calculation:
customers
incoming calls
outgoing calls
etc
Total cost of Coverage MCP
represents the unitised cost for one customer, number of boxes indicates the number of
customers
represents the unitised cost for one incoming minute, number of boxes indicates the
annual number of minutes
represents the unitised cost for one outgoing offnet minute, number of boxes indicates the
annual number of minutes
153
The model
Results
The costs of the coverage network must be recovered (marked-up) from the
services
Equal proportionate
LRIC
Customers
Mark-up

Coverage MCP
Traffic
Outgoing Incoming SMS…
Premium on mobility
LRIC
Coverage MCP
Customers
Mark-up

Traffic
Traffic
Outgoing
Incoming SMS…
The economically optimal method of mark-up utilises
Ramsey pricing economics:

a larger mark-up is applied to services with a lower
elasticity to price change

this is complex and requires knowledge or
assumptions about service elasticity
A number of simpler approaches may be taken, for
example:

LRIC
Customers
Mark-up
Coverage
MCP

Attributable to access
Traffic
Outgoing Incoming SMS…


equal proportionate, as selected by Oftel: markup is applied to all the incremental costs (a proxy
for simplified Ramsey pricing)
‘premium on mobility’: coverage costs are seen
as attributable in equal proportionate terms to
customers and outgoing call minutes
attributable to access: coverage costs are seen as
entirely attributable to customers
NB In all cases the relevant mark-up is calculated and
applied as a percentage increase on the raw incremental
cost of some or all of the services
154
The model
Results
We discuss four key sensitivities
A
Modifying service
routeing factors
D
Reducing the price of
(modern equivalent)
equipment
Base model
Increasing long term
growth in demand
Modifying network
design parameters
C
B
The model
155
Results
Sensitivity – service routeing factors
A
Modifying service
routeing factors
Base model

the economic cost of the equipment required to support demand is allocated to each service in
proportion to the consumption of each resource – new routeing factors will redistribute costs
across the relevant services, and impact the outcome of common cost mark-up
The model
156
Results
Sensitivity – long term demand
Base model

Increasing long term
growth in demand
B
algorithms in the model deploying equipment in advance of future demand would bring forward
deployments - reducing the average utilisation of equipment
The model
157
Results
Sensitivity – network design parameters

network design algorithms would respond to new parameter values, ensuring appropriate
deployments and, for example, impacting the economies of scale present in parts of the network –
impacting the evolution of network utilisation as these economies of scale are exhausted
Base model
Modifying network
design parameters
C
158
The model
Results
Sensitivity – equipment prices
D

Reducing the price of
(modern equivalent)
equipment
Base model
economic depreciation algorithms take into account the expected prices of equipment in the future
– will increase the recovery of costs in earlier years, as the price of equipment in future years is
expected to be lower
159
Executive summary
Introduction to LRIC modelling
Background to the Oftel model
The model:
Cost drivers
Services and increments
Demand forecasts
Network design
Economic depreciation
Network costing
Service costing
Model results
Conclusions
Conclusions
160
The 2001 Oftel model was developed over a long period of time

The Oftel model contains a number of very specific features which have been tailored to meet
the needs of the consultation process in the UK, including a specific variant of economic
depreciation

Iterative processes with Oftel and the industry working group meant that a large number of
complex calculations have been added or refined in the model, often as a reactionary measure
to the demands of the industry working group
Conclusions
161
A number of lessons can be derived from our LRIC modelling experience

Use of a single increment for all traffic is necessary if the model is to be manageable

Understanding how new services will be reflected in the model (and potential corresponding
regulation) should be defined early in the process

Ensuring that the model contains enough fidelity to capture key areas in sufficient detail, yet is
concise enough to be understandable and workable, is critical