MISTELA- AN INTEGRATED SIMULATION MODEL FOR TELEFONICA DE ESPANA
SA,
S. Martinez & A. Barrén ))
Quantitative Methods Department. TELEFONICA DE ESPANA S.A.
Gran Via, 28, planta 12 28013-MADRID (ESPANA)
ABSTRACT’
The main objective of the MISTELA model is to integrate the different aspects of strategic
planning of TELEFONICA DE ESPANA into one single unit. By so doing one is obviously forced to
give up many of the small details in order to be able to look at the larger picture. MISTELA uses a
systemic approach to construct the model described in this paper, Systems Dynamics was chosen,
since this technique permits straightforward combination of different modelling procedures such as
Statistical inference, calibration by trial and error, linear and/or quadratic programming, etc: To give
an idea of the size of the model, it handles about 1,500 equations, definitions and identities. There
are some 700 conceptual variables, and because many of these are vectors, in effect there are about
4,000 scalar variables.
GENERAL STR Fr TELA
Strategic planning is by definition a process. with a long time-scale. At the same time,
planning increasingly implies marshalling a series of complex numerical systems capable of taking
into account the many inter-relationships between the objects being planned. As a result,
simultaneous equation mathematical models are usually appropriate tools to help build the
quantitative. support for this planning. The fundamental idea in a strategic planning model is to
integrate in a single calculation procedure the essence of each different aspect of the planning
process. The model presented here, known as MISTELA2, is made up of twelve sub-models, each a
relatively self contained entity per se. These are:
ol Exogenous variables
02 Prices
03 Demand for access and use
04 Objectives (quantitative business objectives)
05 Construction Investment Required
06 Plant (technology evolution)
07 Personnel
08 Revenue
09 Finance (including Company Accounts)
10 Costs (of Networks and of Services)
ll Financial Results (of Networks and of Services)
12 Investment optimisation.
1 The model has been realized by Martinez, Silvio; Alvarez, Nelson; Barrén, Antonio; Rayego, Pablo;
Gonzélez, German, and Gala, Marta, all of the Quantitative Methods Department of Telefonica de Espafia.
Translation by Graham Johnson.
2 MISTELA stands for "Modelo Integral de Simulacién para Telefonica", that is Integrated Model of
Simulation for Telefonica.
338 SYSTEM DYNAMICS '93
In the following pages a description will be given of the principal sub-models, pointing out
in turn each one's principal interconnections with the others. To begin with, Figure 1 gives a global
overview of MISTELA, showing all the sub-models. The main thing to note here is the large
number of feedback loops between the various parts. In other words, this is a model with a high
degree of interrelationship and strong simultaneity.
The Demand sub-model distinguishes between demand for access (that is, connections),
and demand for use (that is, revenue earning traffic). A typical schematic can be seen in Figure 2-a.
As the figure shows, the relationship between the various access demands (net orders and ceases) is
expressed as two behaviour equations, calibrated by statistical inference from monthly data. These
in turn determine the identity "net demand",
There is a set of two such equations for each Service Group. Generally speaking the
explanatory variables can be grouped into two classes: variables relating to economic activity (GDP,
Industrial Production Index, External trade, Tourism, Telephone density etc.) and prices.
Furthermore, loss of market share is taken into account exogenously as a fraction of the total or
cumulative demand. For each separate Service Group access demand variables are determined in
consultation with areas responsible for the servi -
THE MISTELA MODEL
1
‘Sub-medel
FINANCIAL
RESULTS
EXOGENOUS
VARIABLES Wetworks
ee eew ne eenee- Services
Company oe PLANT
Policy OBJECTIVES Service Groups|] M™'°* NETWORK
- Ends [Service Group: COSTS
- sai , Network Types
OF
= Submodel
: PERSONNEL
: ‘Activity Groups 09 sub-model.
: Service Groups FINANCE
: (including Comp.
i kequis ‘accounts>
‘ ervice Groups
: 106 SERVICE
i cosTS
Seeeeeceneseesseesness: OPTIMISATION Service Types
Service Groupe
Figure 1
SYSTEM DYNAMICS '93. 339
ces, and further ad-hoc studies.
Reliance on expert judgement of this
kind is particularly important when
complete historical data is not
available, for example in the case of
new. products and services. Typical
pairs of equations in this part of the
sub-model are:
: Figure 2.a
NETORDERS = f(prices)* MARKET _sHARE. ....[1] The module (figure 2.b) which deals with
use, that is revenue earning consumption, contains
CEASES = g{ prices) * MARKET _ SHARE... [2] a section dedicated to telephone usage by Service
Group (i.e. Basic Telephony, Packet Switching,
Mobile, Calling Office and Coin Operated Phones
in Private Premises) and another element for Packet Switchitig usage. The first part combines
stochastic equations calibrated using monthly data (local, long distance and international use for
each Service Type)? expressed as number of calls and minutes of calling.
USE DEMAND MODULE
vaniamcee
Figure 2.b
3$ervice Type is a more detailed sub-division of Service Group.
340 SYSTEM DYNAMICS '93
cout "+ SET “translate” objectives into
Figure 3 INVESTMENT = & EQUIPMENT) * PRICESi
” Figure 4 shiows the basic schematic of the Plant sub-model. The basic unit used in this module is the
Service Group. The mathematical relationships used in this sub-model are basically level
(CONNEXt =‘CONNEXt — 1+ NEW _CONNEXt - DECOMt
Figure 4
Schematics of the Personnel sub-model are given as an Employment module (Figure 5) and
a Personnel Cost module (Figure 6) The first of these deals with workforce size, taking into account
4 “Level Equation" is Systems Dynamics terminology for the expression which governs the "level" of stock in
areservoir.
SYSTEM DYNAMICS '93, 341
both what is theoretically necessary and what will actually exist. The second deals with the costs of
that workforce, in terms of manpower and management. The first module's methodology can be
summarised as follows:
PERBONNEL SUB-MODEL
rgonecreee wonstonee tie oeoee.
"Br ncrinee
sore womronce
wonsronce see <—— “aire
Br actinte auarecy
Figure 5
1) Stochastic equations are used to
determine — the theoretical
employment requirements by
Activity Group®, in the absence of
technology improvements. A
‘typical equation is as follows:
(NECE_WORKFORCE™=F(CONNEX,QUAL_OBJ }
2) As the estimating equations
were obtained using cross-section
data, they don't take account of
technological progress. It_ is
therefore. necessary to adjust the
figures for the workforce necessary
using the effects of two sets of
variables: firstly;:the degree of
digitalisation, secondly, in effect
exogenously,-through the use of
time-series tables. This generates adjusted values of workforce necessary to pass to the next stage.
3) Through the use of exogenous equations the total ‘strategically desirable workforce size is
established. Then, the difference (assuming this to be negative) between the strategically desirable
workforce size and the workforce necessary is distributed between the Activity Groups. In the case
where the difference is positive, the strategically desirable employment level is modified downwards.
The second module
calculates the costs of the
workforce, in terms of MANPOWER COSTS = Z REAL WORKFORCE _SIZEk UNTT_WORKPORCE _ COSTS...
manpower and
management. It begins
by calculating workforce
costs according to the
following relation:
0)
wr WORKPORCE CST! = UNTT_RORKPORGE _ Costs ~ 1691 IR®* PRODUC INC...[2]
Were INTR measures. the: effect. of intrinsic wages cost increases (seniority etc.) and
PRODUC_INC captures the effect of salary increases for increased productivity. Management costs
are estimated using the following definition:
‘MGT _COSTSt = MGT _COSTSt ~1* CPI * APLANT * MGT _ COST _ SAVING
Activity Group are a cost accounting sub-division of activities, for example, maintenance, administration, etc.
342
SYSTEM DYNAMICS '93
teaex chOwTH
Prooucrivire +, war maneowen,
"REWARDS.
wrrmoreava seour ‘
‘gavince BUE TO.
|
co a, TET cottor
Wrecnenes| “——> intsteae
_
cost ay acTIVnry
ov
Gout ay acnny <—— “Stee a acta
Tora, MANromen
Decournme
rrreer
REVENUE SUB-MODEL
Figue 6
The fundamental
TEL Mgneaa sa. commen
“ute mseavice ——CMEVEMUE
aks a 4
Ree ae
powesrie .
> eatt Mevenue
ono)
renuarionae ty MrERuariONaL
LS
omen
nEVENE
objective of the Revenue sub-
model (Figure 7) is to calculate
revenues earned by Service Type
and by tariff item (connection
charge, rental charge,
maintenance, equipment sales,
other’.sales; “etc.). Connection
revenues are obtained by
multiplying the net increase in the
number customer lines in ‘service
by the corresponding average
prices; rental revenues from the
Figure 7
provincial, inter provincial and international; for packet
by the. corresponding average prices. In addition, the
product of average number of
revenue-earning connections and
average annual rental charge; call
revenues by multiplying usage (for
telephony: calls in each of the
four categories - _ metropolitan,
switching: Packet Switching Tariff Units)
revenue from sales and maintenance of
terminals is calculated for each Service Group, and also the revenue from radio-paging, maritime
services, and directories.
SYSTEM DYNAMICS '93
There are three parts to the Finance Sub-model: the Profit and Loss Account (with the
same entries as in the company's annual report), the Balance Sheet, and the statement of Source and
Application of Funds (SAF). The normal financial ratios are also calculated. The key points which
govern the operation of this sub-model are:
1) Operating revenue comes from the Revenue sub-model.
2) Costs are dealt with as follows:
Manpower and management costs are derived-from the Personnel sub-model and its associated
costs
© Other operations costs are calculated by varying those of the year before according to the CPI
(Consumer Price Index) and the plant evolution rate (measured as an incremental weighted average
of the connections installed by Service Group). The possibility of increases less than the CPI and/or
the rate of plant evolution is catered for.
3) Further, the value of provisions for exchange losses is given exogenously.
4) Both normal and extraordinary depreciation are taken into account.
5) Lastly, certain variables are considered as. exogenous, such as: financial revenues, charges,
extraordinary revenue, extraordinary expenses, etc.
The Costs of
Networks and. Services sub-
model are defined using the
C_SPHCIFIC = C_PERSONNEL EXTERNAL ,...+C_DEPREC +C_FINAN +C_ADMIN. f) following expressions:
(C_PERSONNEL EXTERNAL ,...c= C_PERSONNEL EXTERNAL ,..1-1*AF_ASSETS ....{2] ;
(C_DEPREC = F- ASSETS *(\/LIFETIME), soo 3] ‘The changes in the
value of fixed assets specific
C_FINAN = TOTAL - FINAN _EXP*/(F_ASSETS )....[4] ch 1 servi
C_ADMIN: = C-ADMIN -*AF_ASSETS ... ] : ee are 7
defining equation of the
usual type, in this case:
F ASSET: = F ASSET; -1+GROSS_CAP_OUT-DEPREC
The cost of “purchasing" from other networks is obtained from the following formal
relation:
C_PURCH_NWKS = SY‘ UNITS_PURCHASED *PRICE,
“Purchases" from services are given by an analogous equation. Networks don't “purchase from
services, so this concept only applies to service costs. Network revenues are calculated according to
the following expression:
NWK REVENUE = S“UNITS_SOLD#*PRICES:
344 SYSTEM DYNAMICS '93
The required margin can be different for each type of network. Clearly, the unit. margin
actually achieved may not be the unit margin desired. The above relation, in words, says that the
“selling price" of circuits of each type of network attempts to reflect the unit cost (networks "sell" -
though it would be better to say "rent" - circuits to other networks and to services). This unit cost is
obtained by, increasing last year's by a cost increase forecast. In time, the margin could become zero
or negative.
Different Service Types sell different types of units. For instance the. Service Access
Network rents circuits, and the Provincial Switched Telephony Network sells minutes of medium-
distance telephone communication. These "sales" are made to customers or external users (that is
clients strictu_sensu) and also to other services, such as Telephone Kiosks and Calling Offices.
Fundamental idea is to. analyse the problem as a square matrix of inter-relationships between
networks and services. Furthermore, the various specific costs mentioned form a matrix of “primary
inputs" with customer demand forming a corresponding vector. There are really two sets of
relations of this kind: the first expresses a physical trade (links leased or "sold", minutes of use sold,
etc.) while the second deals with the financial effects of this trade. These effects are calculated from
tariffs and transfer-prices.as appropriate.
Lastly, the Investment Optimisation sub-model uses as its optimising methodology
quadratic programming, with linear constraints. The squares of the differences of required and
possible investment, and of desired and achievable objectives, are minimised bearing in mind the
relationships which exist between investment and achievement of objectives, expressed as linear or
linearisable equations.
In mathematical terms, using
_ matrix notation:
Min:
1/2*|(/- Ib)" - Ib) |+[U - )|"*](O- Od Where I is the: vector of
. (c Y} 1 )} ( } I¢ y) possible investment. (about 20
Subject_to: elements) generated as.a result of
O=M*I+N quadratic programming. Ip is the
vector of required or desired
investment (calculated by Sub-
model 5), with the same number of elements.as I. O is the vector of achievable or possible
objectives, which is also the result of programming (about 10 or 15 elements). Op is the vector of
desired objectives (calculated by Sub-model 4). A is a square matrix which fits I, which,
subjectively, allows one type of investment to take priority over another.
BUILDING THE MODEL:°
We have selected as the tool to build the model MISTELA the software DYNAMO for
WINDOWS 3.1, because its flexibility to maintain.and add new questions and equations, "builder's
reason", and also for its easy use (friendly interface and integrated environment of windows
software.).
6 Dynamo for windows is copyright of PUGH ROBERTS ASSOCIATES. Windows and MS-DOS are copyright
of MICROSOFT CORPORATION.
SYSTEM DYNAMICS '93. 345
files.
When is running the model, it opens
Sour windows:
The first is decisions, where you can
change the values of variables and build new
scenarios in-a few seconds. The second shows
the results as graphics, and other two as a
numeric display (tables, balance sheet,...).
You can include comments’ and helps about
the model everywhere. i
The Dynamo can’ include saved
scenarios to compare ‘results with the actual
hypothesis: Also you ‘can’ print, save “and
“export screens to others programmes, fax or
‘The requirements to work with MISTELA are:
A personal computer .286/386/486 with 4 (8 better) Mbytes of memory. VGA resolution.
Software: MS-DOS, WINDOWS 3.1, DYNAMO for WINDOWS: , and MISTELA
MODEL.
DECISIONS AND RESULTS
Main of variables which constitute "scenarios" (simulation hypothesis):
1.-Macroeconomic environment
Rate of GDP growth (VIPIB)
Rate of CPI growth (IPC)
Interest rate INTERES)
Industrial Production Index (IPI)
Tourism (TURIS)
External trade index (COMEX)
Exchange rate (CAMBIO): *
2-Tariffs
Connection charges:? (CC) & Rental charges (CA)
Usage-charges (telephony)
zi Metropolitan (PRIM & VMETRO)
- Provincial (PRIM & VPROVI)
- Inter-provincial (PRIM &VNACIO)
- International (PRIM & VINTER)
3.-Quality v
Waiting list, or mean waiting time 7 (DWL=LED)
4.-Competitive environment
Market shares of connections 7 (PCMPN)
Market sahres of _ revenue-earning
(ANALY0,...ANALY3)
7 By Service Group.
traffic (domestic and international)
348 SYSTEM DYNAMICS '93.
5.-Business Actions
Saving in investment through improvements i in management (INVEMAE)
Savings in‘external purchasing and supplies (EDESL & FCOGO) .._
1.-Markets
‘Demand for conncetions 7 (@N).
Demand for use 7 (CONAGS, COINGS)
Waiting list 7 (WL=LE)
New connections 7 (NEW=ALTAS)
2.-Plant evolution
Installed lines 7 (WL=LI)
Lines in service 7 (LS)
Revenue-cearning lines in service 7 (RFL=LF)
Lines of spare capacity 7 (SL=VAC)
Degree of digitalisation (DD=GD)
3.-Jnvestment necessary by programme (PRGXX)
4.- Workforce size by activity type (EMPREAL)
5.- Profit & loss account, balance sheet and S.A.F.
6.- Cost accounting by network and service (under revision).
REFERENCES.
ARACIL, J. 1986, Introduccién a la dindmica de sistemas. Alianza Editorial, Madrid.
MARTINEZ, S.; and REQUENA, A. 1986. Dindmica de Sistemas: 1. Simulacién por ordenador.
Alianza Editorial. Madrid.
MARTINEZ, S.; and REQUENA, A. 1986, Dindmica de Sistemas: 2. Modelos. Alianza Editorial.
Madrid.
THEIL, H. 1964. Optimal Decision Rules for Government and Industry. North-Holland.
Amsterdam.
AUTHOR'S BIOGRAPHY.-
Martinez, Silvio, Director del Dto. Métodos Cuantitativos de Telefonica de Espafia S.A. Gran Via
28, pl, 12. 28013-Madrid (Spain).Fax 34-1-5232350.
Education: PhD. Economic Agricultural Engineering and PhD in Economics. Professor in Consejo
Superior de Investigaciones Cientificas in Madrid. Lecturer in Universidad Aut6noma de Madrid
and consultant in Regional and Agricultural. economics.
SYSTEM DYNAMICS '93 37
At present work in System Dynamics Modeling for Strategic Planning.
He is the author of the three books on System Dynamics, and of several other books on Regional
and Agricultural Planning.
Barrén, Antonio, Assessor on informatic & telecommunications areas.
Superior Engineer of Telecommunications (E.1.S.T. of Madrid).
Experience in a few Areas of TELEFONICA DE ESPANA S.A. during 15 years: Specificator,
informatic consultant and support, management, new services designer, documentation writer.
Member in the C.C.L.T.T., in Normalization groups. Inner teacher for the telephon workers.
Reference: SDC93.DOC
348 SYSTEM DYNAMICS '93.
CC BY-NC-SA 4.0