An airline number of daily flights on a route, peak hours at airports, waves at hub airports, maximum daily flights to be offered to the customers on a route, number of aircraft different types in a airline fleet, … this is a set of air transport issues which are strongly related to each others and which are the direct or indirect consequence of the daily distribution of (business) demand on city pairs and how an airline tries to capture most of it. Statistical analysis and modelling cannot provide efficient enough tools for assessing the actual weight of the sole scheduling/frequency factor on the passenger choice as, for instance, the fares and the frequent flyer programmes do play a significant role. This is why we favour a behavioural modelling which would be based on the combined attractiveness of the local departure and arrival times of a flight together with the bell shaped curve of the flight attractiveness over the day since the further the actual departure time from the desired one the lower the number of people willing to take it. We expect this approach to answer the following questions: 1/ what is the maximum number of daily flights beyond which it does not pay off adding a new one? 2/ when these flights should be operated over the day? 3/ how the flight programme of 2 and more contenders affects their related market shares? 4/ how this approach compares/complements with the S-curve theory of de Neufville?

1. Daniel SALLIER 1
DAILY DEMAND DISTRIBUTION & FLIGHT PROGRAMME ATTRACTIVENESS
FOR THE PASSENGERS ON A CITY PAIR
By Daniel SALLIER
Aéroports de Paris
Bât. 530 – Zone Orlytech
9, Allée Hélène Boucher
Orly Sud 103
94396 Orly Aérogare cedex
France
Telephone: +33 6 82 84 12 56
daniel.sallier@adp.fr

2. Daniel SALLIER 2
ABSTRACT
The daily number of flights operated by an airline on a route, peak hours at airports, waves
amplitude at hub airports, maximum daily flights to be offered to the customers on a route, number
of aircraft different types in a airline fleet, … this is a set of air transport issues which are strongly
related to each other and which are the direct or indirect consequence of the daily distribution of
(business) demand on city pairs and how an airline manages to capture most of it.
Statistical analysis and modelling cannot provide efficient enough tools for assessing the actual
weight of the sole scheduling/frequency factor on the passenger choice as, for instance, the fares
and the frequent flyer programmes do play a significant role. This is why we favour a behavioural
modelling which would be based on the combined attractiveness of the local departure and arrival
times of a flight together with the bell shaped curve of the flight attractiveness over the day since
the further the actual departure time from the desired one the lower the number of people willing to
take it.
This approach provides clues as to:
1/ what is the maximum number of daily flights beyond which it does not pay off adding a new
one?
2/ when these flights should be operated over the day?
3/ how the flight programme of 2 and more contenders affects their related market shares?
4/ how this approach compares/complements with the S-curve theory of de Neufville, Lenoir
and other prominent specialists?
Keywords:
Demand, Forecast, Modelling, Frequency, Scheduling, Market Share

3. Daniel SALLIER 3
1. INTRODUCTION
Which daily number of flights an airline has to operate on a route?
This is a very simple question to ask of which the answer has a large set of tremendous
consequences. Consequences for the airline in terms of flight programme, operational options, fleet
planning, commercial competition. Consequences for the airports the airline is operating from in
terms of peak hours developments, terminal dimensioning, number of runways, area of aprons,
noise restrictions, etc… Consequence for the aircraft manufacturers which will have to supply the
good set of aircraft types which can meet the airlines requirement while manufacturing enough
aircraft of each type to get its return on the project. Consequence for the ATC organisations which
will have to handle the traffic the airlines are putting to the air. Consequences for the leasing
companies which have to build a portfolio of aircraft types to be highly praised by the airlines so
that they secure the residual value of their assets. Consequences for the bankers and the credit
export organisations which will have to finance aircraft acquisitions and get sure that the aircraft
they are financing will have a good residual value.
The reason why such a question has such dramatic consequences is quite simple: a given volume of
passengers to carry out translates into a given volume of seats to supply the market with. The same
volume of seats can be achieved either by operating rather big aircraft at a rather low level of daily
flights or by operating smaller aircraft at a higher level of frequencies or any solution in between.
Unfortunately the level of daily flight affects the business and high yield oriented demand an airline
can tap into while at the same time the smaller the aircraft the higher its operating cost per seat.
Hereafter is a "simplified" causal diagram1
on how these different elements affect each other:
Figure 1: Simplified causal diagram
1
Causality is defined here in its classical meaning in physics or system dynamics and does not refer to any Granger
causality test of any sort.

4. Daniel SALLIER 4
The red boxes and lines are the elements and the relations this paper is mostly focused on.
Because the passenger demand on a route an airline can capture is not only a matter of price,
frequencies and departure/arrival times. Because it is a matter of brand, of service offered to the
passenger, of frequent flyer programme, of competition with other airlines and even of the other
itineraries offered to the market. Because most of the time we are dealing with small if not
insignificant numbers of flights2
. Because of all these reasons the use of statistical tools does not
prove that efficient for identifying the effect of this or that specific and well identified phenomenon
on the demand with an acceptable risk of being inaccurate or even wrong.
These are the very reasons why we have favoured a behavioural approach and modelling.
To start with we will see how it is possible to turn departure and arrival time attractiveness for the
passengers into a daily distribution curve of demand.
In a second time we will introduce the concept of flight attractiveness over the day function. We
will generalise the flight attractiveness function to the flight programme attractiveness one. By
combining the daily demand distribution curve on the one hand and the (programme) flight
attractiveness function on the other we introduce the concept of daily demand coverage of a flight
or a flight programme and, as a consequence, the concept of optimal demand coverage which is the
maximum coverage a flight programme can deliver.
In a third time we will look at the demand coverage function properties, among which we will pay a
special attention to:
1/ the evolution of the flight programme (demand) optimal coverage as a function of the number
of daily flight considered;
2/ the market share estimate between two and more contenders on the same route based on the
number of flights and their departure/arrival times.
In a fourth time we will look at the approach sensitivity to the various parameters which are used.
The fifth and last part is the conclusion after a short discussion leading to future additional
developments.
2. DAILY DEMAND DISTRIBUTION
It could be very attractive for a long haul flight business passenger to arrive at 5H00 am for it gives
him the opportunity to have a taxi and go to his hotel before morning traffic jams develop, then to
have a shower before attending his first morning meeting.
2
One daily return flight on long haul destination can be considered for being far from a standard situation. For
instance, in Paris, over 313 (airline, airport) pairs served in 2010 which are distant of 5,500 km (2,970 NM) and
more from Paris, only 1 is served 4 and more times daily, 8 are served 2 to 2.99 times daily, 26 are served 1 to 1.99
time daily.

5. Daniel SALLIER 5
On the other hand a 5H00 am arrival time is far from being attractive for a short haul business
passenger after a 1H30 flight for it means that he would have to wake up at 2H00 am at the latest to
go at the airport and board his very early flight and wait three to four hours before having his
morning meeting.
This very simple examples tells us that, depending on the type of trip, we can credit each hour of
the day considered as a departure time or an arrival time with an attractiveness index: long haul
flight arrival at 5H00 am = very attractive, short haul flight arrival at 5H00 am = not attractive at
all.
In our opinion a set of 4 attractive indexes is far sufficient for qualifying each hour of the day:
N Not attractive
S Slightly attractive
A Attractive
V Very attractive
The following tables represent a set of such hour related attractiveness indexes:
Table 1: Example of hour attractiveness tables

6. Daniel SALLIER 6
This is a common sense based set of tables of which the attractiveness index would have to be
confirmed by passenger surveys bearing in mind that local habits may differ from one country to the
other3
.
Quantitative modelling cannot deal with qualitative index such as "Not attractive departure time". It
is a problem which can be overcome very easily by attaching a value to each qualitative index:
Table 2: value sets of the hours attractiveness indexes
In addition the tables refer to flight / trip categories such as short, medium, long and very long haul
which should be categorised too. The following table illustrates the flight categorisation which can
be used:
Table 3: Flight categories
Schedule specific attractiveness index
At this stage we can define the schedule specific attractiveness by combining its local departure and
local arrival time attractiveness indexes. In addition we have to consider the possibility that
departure and arrival times are not equally valued by passengers, … so let us introduce:
 0,1  , the departure time weight
DI the attractiveness index of the a given local departure time
AI the related attractiveness index of the corresponding local arrival time
SI the related intrinsic attractiveness index of the corresponding flight
They are 2 ways to define the schedule intrinsic attractiveness index SI :
3
One obvious example, at least in Europe, is the working hours in Spain which differ significantly from those of the
rest of western European countries.

7. Daniel SALLIER 7
1/ the arithmetical approach:  1S D AI I      I if none of the index refers to the "not
attractive" category, otherwise.0sI 
2/ the geometrical approach:  1
S D AI I I
 
  if none of the index refers to the "not attractive"
category, otherwise.0SI 
Cross arrival/departure attractiveness index
The issue with this schedule attractiveness calculation is that we can determine a schedule as being
attractive while it is not and the other way around.
For instance a long haul flight which would take off at night and arrive at night too can be identified
as being highly attractive while it is not4
. Another example are rather short distance long haul
flights or pretty distant medium haul flights taking off quite late at night for an early arrival which
should be theoretically qualified as being very attractive while in fact they are not for the passengers
have almost no time left to rest during the cruise time5
.
A way to overcome the problem is to introduce cross arrival/departure attractiveness tables as
illustrated hereafter:
Table 4: Departure/arrival cross-attractiveness
Once again passenger surveys would have to be used for index validation purposes. The upper
tables are using different periods of the day such as morning, mid-day, late-day and night which
should be specified as done hereafter:
Table 5: Categories of day periods
4
This is the case of late departures from Europe for south-east Asian destinations.
5
This is the case of late departures from the Arabic peninsula to Europe.

8. Daniel SALLIER 8
Schedule (composite) attractiveness index
At this stage we can define the composite attractiveness of a specific schedule by combining its
intrinsic attractiveness and its cross arrival / departure attractiveness indexes. In addition we have to
consider the possibility that the intrinsic and cross attractiveness are not equally valued by
passengers, … so let us introduce:
 0,1  , the schedule intrinsic attractiveness weight
SI the intrinsic attractiveness index of a given schedule
CI the related cross attractiveness index of the corresponding schedule
I the related (composite) attractiveness index of the corresponding schedule
Once again they are 2 ways to define the schedule (composite) attractiveness index :I
1/ the arithmetical approach:  1S CI I     I if none of the index refers to the "not
attractive" category, otherwise.0I 
2/ the geometrical approach:  1
S CI I I
 
  if none of the index refers to the "not attractive"
category, otherwise.0I 
By varying the departure time from 0H00 am to 11H59 pm it is possible to get the evolution of the
schedule (composite) attractiveness index over the day.
In the following example we consider a 8H00 trip time westward, -6H00 time difference, 0.5  ,
0.8  and a  hour attractiveness set:0,1,2,3
Figure 2: Example of (composite) schedule attractiveness

9. Daniel SALLIER 9
 t for a flight which would depart at t, is likelyWe can assume that the relative hourly demand
to be proportional to its related schedule composite attractiveness:
 
 
 
23:59
0:00
I t
t
I t dt
 

so that
23:59
et us call
 0:00
100%t dt   .
 tL the relative demand density function which is expressed in percent of total daily
ur.
he schedule attractiveness curve of the upper example translates into the following relative daily
traffic per ho
T
demand distribution curve:
Figure 3: Example of daily demand distribution
is a demand curve which does not take into consideration any actual flight supply. At this stage
3. FLIGHT ATTRACTIVENESS & DEMAND COVERAGE
et us now suppose that we would like to take off at 8H00 am sharp for a short 1H30 flight and that
we are offered such a departure time. 100% of the people who would like to take-off at 8H00 are
It
we already start to get a better understanding of the reasons why airports and airlines are faced with
peaks of departures and arrivals over the day which just reflects passenger needs.
L

10. Daniel SALLIER 10
happy with the schedule they are offered. May be 99.9999% of the people who would prefer to
depart à 7H55 would be happy with an 8H00 am departure. It is more than likely that almost 100%
of the people who would definitely prefer to take-off at 4H00 pm are not happy with the morning
departure they are offered.
The former example tells us that, for a given flight leaving at a given departure time, the farther the
esired departure time from the actual departure time, the lower the number of "happy customers".d
This relation can be translated into a flight attractiveness curve which, for a given actual departure
time, provides the percentage of the instantaneous demand which keeps being interested by this
flight as a function of the difference between the actual departure time and the desired one. Such a
function can have different shapes: triangle, rectangular, trapezoidal, bell curve for the more
common ones:
Figure 4: Example of flight attractiveness curves
We tend to favour bell-s ions:haped functions, and more precisely Gauss-like funct
 
2
Dt t 

0.6
FlI t e
  
 where
 FlI t is the flight attra s function
is the tim
ctivenes
t e
Dt the actual departure time
 0.5 50%Fl DI t   is the flight attractiveness span so that
nd
2
tripT
  where is the trip / flight time.A tripT

11. Daniel SALLIER 11
This value of the flight attractiveness span echoes an old rule of
states that the maxim
the thumb in the industry which
um number of daily flights to a destination is equal to 24 divided by the trip
time.
In the case of multiple daily flights operated by an airline the flight programme attractiveness
 I t is defined as the envelop of the set of individual flight attractiveness curves:FP
 max iFl  1
FP
i toN
I t

I where N is the total number of daily flights
emand coverage of a flight
verage of a flight or a flight programme as the part of the daily
terested by this flight or flight programme. Demand coverage is
C is the demand coverage of the flight programme
D
We can define the demand co
emand which keeps being ind
expressed as a percentage of the daily demand.
   
23:59
C I t t dt   where6
0:00
FP
 FPI t is the flight programme attractiveness function
 t the relative demand density function
The rizes the different concepts exposed in this chapter, in
hich the dark blue area represents the programme flight (demand) coverage.
following chart illustrates and summa
w
Figure 5: example of daily demand, flight programme and demand coverage
6
This definition echoes the kernel techniques used in statistics where FPI would be the kernel function

12. Daniel SALLIER 12
It is quite obvious from the upper chart that the demand coverage of a flight or a flight programme
fully depends on when the flight or the set of flights are departing. This is illustrated in the
following chart which represents the demand coverage as a function of the departure time.
Figure 6: Example of demand coverage as a function of departure time
The best departure time, no competition assumed, which would maximise the demand coverage
would be at about 18H30 pm and would satisfy about 38% of the daily demand, the second best
choice would be a departure at 10H30 am which would satisfy about 33% of the daily demand.
4. DEMAND COVERAGE PROPERTIES & MARKET SHARE CALCULATION
Flight programme and demand coverage maximisation
The demand coverage concept leads very naturally to 2 complementary questions:
1/ for a given number of daily flights, what is the flight programme characterised by its set of
departure times which maximise the demand coverage? We call this maximised coverage
value the optimal coverage.
2/ How the optimal coverage evolves as a function of the daily number of flights?
ecause we ptimisation
lgorithms. In addition, we can end up with significantly different flight programmes resulting in
the same or pretty close optimal demand coverage; there is no unique solution rule applying here.
B
a
are dealing with non linear relationships, we cannot use classical linear o

13. Daniel SALLIER 13
The following chart which is based on the Paris to New York non-stop route shows how the optimal
coverage evolves as a function of the daily number of flights.
Figure 7: Example of optimal demand coverage vs daily no. of flights
Our own experience is that the slope of the optimal coverage curves smoothes down in the close
vicinity of 80% to 90% optimal demand coverage. It means that adding a new service to a flight
programme which already accounts for 80% to 90% demand coverage has but a marginal capability
to capture additional demand.
This is the reason why we consider that 80% to 90% optimal coverage is the adequate criteria to be
used for determining the maximum number of daily flights it is worth operating on a route from a
commercial point of view.
Because of the time difference between cities located at different and distant places, the daily
demand distribution pattern fully depends on the flight origin and destination of which the
consequence is that the return flight demand pattern tends to be quite different from the one of the
outward trip. This is the reason why outward and return routes may differ in the maximum number
of daily flights which are attractive for the passengers.
Furthermore, the by-product of the optimal coverage function determination is the departure /
arrival times of the different optimal flight programmes. Given a set of destinations and the airlines
which serve them for an airport, given a network for an airline this approach allows to get a better
understandin lop over the
ears.
g on how the peak hours / the waves in a hub and spoke system can deve
y

14. Daniel SALLIER 14
Market share
We assume here that all the airlines operating the same route differ only by their flight programme,
ut have the same fare, the same frequent flier programme advantages, the same travelling
tantaneous" number of passengers identified by their desired departure
me t the airline can capture is proportional to the airline coverage curve value for t over the sum of
b
agreements with corporations, the same brand fame, the same quality of service, the same ...
We can assume that the "ins
ti
coverage curves of all the competitors for t. If we sum up this number of "instantaneous" passengers
along the complete day over the total number of daily passenger we get the airline market share:
 
 
 
23:59
,FP i
i
I t
S t dt   ,0:00 FP i
i
iS is the market share of airline i;
 ,FP i
I t where
I t is the flight programme attractiveness function of the airline i;
 t the relative demand density function
ne can easily checks that 1iS O
i
The following example is based on a 7H55 trip time from Paris to New York, -6H00 time difference
nd 2 airlines only competing.a
0 1 2 3 4 5 6
0 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1 100.0% 50.0% 23.3% 23.3% 19.2% 18.3% 18.3%
2 100.0% 76.7% 50.0% 42.3% 38.2% 37.4% 37.0%
0.0% 76.7% 57.7% 50.0% 45.9% 45.1% 44.3%
.9% 48.1%
.7% 62.6% 54.9% 51.1% 50.0% 49.2%
6 100.0% 81.7% 63.0% 55.7% 51.9% 50.8% 50.0%
#2
Daily no. of
flights
Airline #1
High yield market share of ariline #2
3 10
line
4 100.0% 80.8% 61.8% 54.1% 50.0% 48
Air
5 100.0% 81
Table 6: Market share example
his table tells us that once it reaches a programme of 4 daily flights the airline #1 cannot preventT
any more the airline #2 from building its market share.

15. Daniel SALLIER 15
Based on the table data, let us assume, for instance, that airline #1 is operating 3 daily flights and
airline #2, twice only. Airline #2 market share is 42%. If airline #1 adds a new daily flight and goes
to 4 daily, it can reduce #2's market share to 38%. If airline #1 where to go to 5 and even 6 daily
flight, this would not significantly affect airline #2 market share any more.
Frequency policy is always aimed at taking or defending airline's share of the high yield market:
first, business and premium economy class or up to 4% to 8% of the economy class. High yield
market accounts for 15% to 20% of the total number of passe f a Paris to New York route for
instance.
Let us assumed, still based on the former table, that the 2 airlines are offering 4 daily flights each. If
one of the 2 contenders adds a new daily service it would gain 1.1% of the high yield share which
represents 2.1% of its own high yield market and 0.3% of its tal traffic while going from 4 to
5 daily flights means increasing its own seats supply to the by 25% if it were to keep the
same aircraft type!
ow let us assume that airline #1 is already offering 4 daily flights. If the airline #2 is operating
ld bring the airline #2 with 20% more high
ield passengers and 3% only of its total traffic. Airline #2 can but shrink its aircraft size and coach
ge criteria for assessing
e maximum number of daily flights it is economically worth operating on a route.
an S-curve distribution.
ngers o
own to
market
N
twice a day and want to go to 3 daily flights, it means adding 50% more seats if the same aircraft
type where to be used, while adding this new flight cou
y
class fares for having a chance to get a decent load factor on its 3 daily flights operations ... the
economical return of adding this new flight is pretty far from being granted.
Market share based analysis tend to confirm the 80% to 90% optimal covera
th
Based on the former table, the following chart represents the share of airline #2 high yield market as
a function of its share of the daily number of flights. According to Nathalie Lenoir's theory7
we
should get
7
Lenoir N, "Air Transport Network", ENAC, Aviation Economics and Econometrics Laboratory (Toulouse)
Page 53 of her presentation, Mrs. Lenoir proposes the following relation:

16. Daniel SALLIER 16
Figure 8: Demand market share vs frequency market share
Our approach leads to an S-curve like distribution too, which would have to be confirmed by a
higher number of cases yet to be studied. The issue is that the shape of the S-curve we get differs
significantly from the empirical one, mostly at the inflexion point where ours is almost flat while
Mrs. Lenoir's one
We are more "Lenoir-conform" if we consider how airline #2 market share changes as a function of
its number of daily flights while keeping constant the number of flights operated by airline #1.
Th ates 2 daily flights and airline #2
operates 0 to 6 daily flights.
is almost vertical.
e following chart is based on assuming that airline #1 oper

17. Daniel SALLIER 17
Figure 9: Traffic market share vs frequency market share
In fact the 2 curves cannot be directly compared since Mrs. Lenoir does not filter out phenomenon
which are not strictly frequency and scheduled related.
5. SENSITIVITY ANALYSIS
We will illustrate the method sensitivity to its various parameters on the base of the Paris to New
York route: 7H55 flight time, -6H00 time difference between Paris and New York, α = 0.5, β =0.8
and  hour
ry attractive" into corresponding numbers.
he following chart shows us how ffected by a change in the set of
alues used for the attractiveness
0,1,2,3 attractiveness index set.
Sensitivity to index attractiveness set of values
e had to convert qualitative attractiveness such as "VeW
T the daily distribution curve is a
index:v

18. Daniel SALLIER 18
Figure 10: Sensitivity to the attractiveness index set
Of course changes in demand distribution
curves, but to a very "reasonable" extent.
ensitivity to α parameter
e weight in the schedule specific attractiveness calculation. The higher
e α value the higher the weight of the departure time index and the lower the one of the arrival
the set of parameters translate into changes in the daily
S
α represents the departure tim
th
time in the schedule attractiveness calculation:
Figure 11: Alpha sensitivity

19. Daniel SALLIER 19
Once again the approach seems to be pretty robust and does not show any "pathological" behaviour.
Sensitivity to β parameter
β represents the schedule specific attractiveness weight in the schedule composite attractiveness
calculation. The higher the β value the higher the specific attractiveness and the lower the cross
attractiveness in the schedule composite attractiveness calculation:
Figure 12: Sensitivity to beta parameter
Once again the approach seems to be pretty robust and does not show any "pathological" behaviour.
Sensitivity to calculation mode
The combination of attractiveness indexes or curves can be done either by using arithmetical
barycentre calculation such as  1B X     Y or geometrical barycentre calculation such as
 1
B X Y  
 
It looks like the calculation mode does not significantly affect the shape of the daily demand
distribution curve; geometrical based calculation tends to increase the "picture contrast":

20. Daniel SALLIER 20
Figure 13: Calculation sensitivity
Sensitivity to flight attractiveness span parameter
The following chart is based on the Paris to New York case assuming a 7H55, almost 8H00 trip
time leading to consider as a base case a 4H00 span for the flight attractiveness bell-shape curve.
We have been considering a span value of 2H30 up to 5H30 and look at how the daily demand
coverage is affected.
Figure 14: sensitivity to flight attractiveness span
In our opinion the approach detailed here is by far the most sensitive to the flight attractiveness span
parameter. We have not yet done the sensitivity analysis up to the optimal coverage as a function of
the daily number of flights since it would require the development of an automatic coverage
optimisation algorithm we have not done yet.

21. Daniel SALLIER 21
6. DISCUSSION AND CONCLUSION
The sensitivity analysis suggests that the approach detailed in this paper is pretty robust and for so
reflects very likely the actual passenger behaviour. Nevertheless, they are 2 points this approach
does not take into consideration:
- The desired return flight availability which may affect the passenger outward flight selection.
For instance an airline operating a daily morning flight on a short haul route cannot satisfy those
among its customers who want to go back home at night. On long haul destination the similar
concern is for less than daily flight programmes. It means that the approach exposed in this paper
can but over estimate the market share of low frequency flight programmes (less than daily on
long haul routes, less than twice daily on short / medium haul routes). One way to address this
issue is to split the daily demand into journey categories such as same day return demand (i.e.:
28% of the daily demand), next day return demand (i.e.: 52%), etc... Only passenger surveys can
provide the demand
hardly to be met in the competition, for instance,
between a non-stop flight and a connecting flight in some hub. This trip time difference question
turns out to be much more com e have already started a
very promising research w a very important subject to
g a single value for each departure (arrival) time which maximises
e
a far
o
Th
de
-
ort pair;
Th
percentage attached to each journey categories;
- Difference in trip time between the different airlines serving the same city pair or in the
competition between air travel and high speed train.
In this approach we have assumed that all the contenders on the same route have a very close if
not identical trip time. It is an assumption
plex to address than the former one, but w
ork on this issue. Useless to mention it is
examine very closely for it would provide a better understanding on the share of the market by
different hub airports, the share of the market between non-stop service and "hubed" ones, etc.
In addition, instead of determinin
th demand coverage, it is worth considering a slot of departure (arrival) times which almost
ximise the demand coverage (i.e.: optimal coverage minus 2% as a slot criteria) this would bem
m re realistic.
ose are avenues for additional researches to come. The approach we have detailed in this paper is
livering what it has been designed for:
the ability to provide the maximum number of daily flights it is worth operating on an airport
pair by an airline;
what are the optimal departing / arrival times of the flights on an airp-
- how frequencies and departure / arrival times affect the share of the market of each competitor
on an airport pair.
e scope of applications of this approach is quite large:

22. Daniel SALLIER 22
-
"tactical" scheduling: at what time one has to schedule his departing flights over the day in order
1/ the O&D demand from spoke airport #1 to hub;
o airport #2.
Each of the 3 categories of passengers demanding their scheduling needs to be fulfilled;
- ng of daily peak hours for an airport;
ival slots;
the airlines and/or the airports;
(A380) directorate with the explicit aim of getting a better estimate of the maximum number of
aily flights it is worth operating on a route for it directly dictates the aircraft seating capacity
Pa terest on this topic for a large airports have to build long term
recast on ATMs by aircraft type for its long term developments (master plans). Is not a terminal
sin ets? Environmental, noise related issues, may require the need for an airport to get a
bet evolve on a long term future for it may be
fac
7. REFERENCES
elobaba P, Odoni A, Barnhart C, "The Global Airline Industry", John Wiley & Sons (2009)
ecember
004)
ca usiness Administration, Harvard University
oston) (1972)
, "Air Transport Network", ENAC, Aviation Economics and Econometrics Laboratory
oulouse) (March 2010)
fleet planning for which one of the pending question is the trade-off between frequencies and
capacity;
-
to maximise his market share;
- more commercially efficient hub scheduling for we are dealing with 2 flights but 3 sorts of
demand:
2/ the O&D demand from hub to airport #2;
3/ the O&D demand from airport #1 t
long term forecasti
- providing rational arguments in airport / airline negotiations on departure / arr
- provide Civil Aviation Authorities with arguments to oppose on frequency related
issues/disputes with
- etc.
My initial R&D works on this topic took place 10 years ago during my time in Airbus at the A3XX
d
which will be required. Few years later as chief of the stats and forecasting group of Aéroports de
ris, I re-launched my R&D in
fo
designed to accommodate A380 types of aircraft quite different from those designed for regional or
gle aisle j
ter understanding on how its noise print is likely to
ed with legal noise constraints8
.
B
Cheung J, "Airline Competition and Operating Strategy in Multi-Airport Systems", MIT (D
2
Fruhan W, " The fight for competitive advantage: A study of the United States domestic trunk air
rriers", Division of Research, Graduate School of B
(B
Lenoir N
(T
de Neufville R and Odoni, A. "Airport Systems Planning, Design, and Management", McGraw-Hill,
New York, NY. (2003)
8
This is the case of Paris - CDG for instance of which the annual noise energy of its traffic should not exceed the
average value of its 1999, 2000 and 2001 noise energy levels.