Integrated Public Transport Planning: Data, Modelling, Operations and Economic and Financial Viability

2. Before designing the network, mobility must be understood

One of the most common methodological errors is to begin the analysis by looking exclusively at the public transport system itself: passengers by route, trips operated, vehicle-kilometres, frequencies, occupancy levels, regularity or commercial speed. All this information is essential, but it fundamentally describes what is happening within the network, not necessarily what is happening in the city.

Suppose, for example, that a route is losing passengers. Ticketing data may confirm the reduction, the operations management system may show whether regularity has deteriorated, and GPS records may identify a possible fall in commercial speed. However, none of these factors may actually explain the problem. Residential patterns may have changed, an employment centre may have relocated, a new urban area may have been developed, or new origin-destination relationships may have emerged that the current network configuration no longer serves adequately.

There is, therefore, a fundamental difference between studying a bus network and studying mobility. The first approach allows us to understand how the service is performing, while the second attempts to understand why trips are made, where they take place and what role each transport mode plays within them.

Planning should begin with this second question. We should first ask how the city actually moves and, based on that understanding, determine what role public transport currently plays and what role it could play in the future.

3. Surveys remain useful, but they should no longer support the entire demand model on their own

For decades, mobility surveys and origin-destination surveys have been among the main tools used in transport planning, and they remain valuable in many studies. They provide information that is difficult to obtain from automated data sources, such as trip purpose, car availability, perceptions of the service, certain socioeconomic characteristics or users' stated preferences.

The limitation appears when we attempt to reconstruct the entire mobility pattern of a city exclusively from surveys. A survey necessarily works with a sample and, however carefully selected, that sample represents only a proportion of actual trips. In addition, responses are not always fully consistent with reality and may be affected by different forms of bias.

Surveys are also costly and complex to undertake, which normally prevents large-scale campaigns from being repeated frequently. Meanwhile, mobility changes not only from year to year, but also between months, between weekdays and weekends, and between different periods of the same day.

The enormous availability of observed data today therefore represents a particularly important opportunity for transport engineering. We can now work with information derived from mobile phone data, electronic ticketing, operational systems, GPS, sensors and automated counting campaigns. The methodological shift, however, should not simply consist of replacing surveys with Big Data, but rather of combining different sources to build a representation of demand that is more robust than that provided by any single source.

4. There is no perfect data source

Mobility data derived from mobile phones provide an extraordinary capacity to observe the territorial dimension of mobility. They allow us to identify relationships between zones, temporal patterns, and urban and metropolitan movements across volumes of population that are extremely difficult to achieve through conventional survey campaigns.

In Spain, the availability of large-scale mobility datasets derived from mobile phone data promoted by the Ministry of Transport and Sustainable Mobility has opened up particularly interesting possibilities for the construction and validation of origin-destination matrices. However, having a very large number of observations does not automatically mean having all the information required to characterise the operation of a particular public transport network.

Ticketing provides a second layer of knowledge by allowing us to analyse actual system use, validations and the temporal distribution of demand. Operations management systems and GPS provide the operational dimension —actual vehicle positions, journey times, regularity, commercial speed and trip completion— while specific boarding and alighting counts make it possible to reconstruct passenger loads along routes and verify where passenger movements actually occur.

Traffic counts, finally, provide information on the part of demand using the road network. Each source therefore observes a different part of the same phenomenon. A particularly robust methodology may consequently be based on a combination of mobile phone data, ticketing, operational/GPS data, boarding and alighting counts, traffic counts, and territorial and socioeconomic information.



What matters is not accumulating databases, but cross-checking them. If several independent sources point towards the same mobility structure, confidence in our interpretation increases considerably. If they do not coincide, that divergence should not be hidden through automatic adjustments, but investigated, because it probably indicates that there is still part of the system's behaviour that we have not properly understood.

5. The origin-destination matrix is one of the most important assets of any model

Formally, an origin-destination matrix is simple: each cell represents the number of trips made between an origin zone and a destination zone during a given period. Its construction, however, is one of the most delicate parts of any transport model because the matrix is, in reality, a mathematical representation of the mobility structure of a territory.

Practically every subsequent decision depends on it. A poorly constructed matrix may even produce a model that appears to perform correctly, since certain parameters can be adjusted until the assignment reproduces observed traffic counts at selected points reasonably well. This does not guarantee, however, that the origin-destination relationships generating those flows are the correct ones.

The problem becomes particularly relevant when the model is used to analyse the future. Two different matrices may reproduce relatively similar traffic volumes under existing conditions and react completely differently when a new infrastructure, public transport route, urban development, traffic restriction or fare modification is introduced.

For this reason, model calibration should not simply mean making simulated flows resemble observed flows.



Demand must also be consistent with the other available sources —mobile phone data, ticketing, counts, operational data and territorial information— and true robustness is achieved when all these sources provide a reasonably compatible description of the system.

6. There is not just one matrix either

Referring to “the mobility matrix” may be an excessive simplification, because urban mobility is not static. The travel structure at 08:00 may be significantly different from that at 14:00, just as origin-destination relationships during a peak hour may differ from those observed during off-peak periods or at weekends.

A sufficiently detailed planning exercise may require different matrices according to time period, type of day, trip purpose, population segment or mode of transport. This temporal dimension is particularly important in public transport, since the network is not designed simply to accommodate a given daily demand, but demand that is simultaneously located in space and time.

This temporal distribution also has a direct economic consequence. Highly concentrated demand may require a large amount of resources for a very short period, whereas demand that is more evenly distributed throughout the day may allow the same fleet to be used considerably more efficiently.



From both a planning and operational perspective, therefore, the shape of demand can be almost as important as its total volume.

7. Macrosimulation: moving from observing the system to experimenting with it

Up to this point, we have been trying to understand what is happening, but planning needs to answer a different question: what will happen if we change some of the conditions of the system? This is where strategic modelling, or macrosimulation, comes into play.

A strategic model should not be understood simply as a sophisticated graphical representation of a road network accompanied by a set of public transport routes. Its real value lies in mathematically representing the relationships between demand, supply, travel times, costs and user decisions, allowing us to experiment with the system before physically modifying it.

One of the most widely used conceptual structures remains the classic four-stage model: trip generation and attraction, distribution, mode choice and assignment. The fact that this methodology has been used for several decades does not mean that it has lost its usefulness. On the contrary, the availability of new sources of information now allows each of these stages to be informed and validated using much broader and more accurate observations.



Macrosimulation ultimately allows us to move beyond using data simply to describe the present and start using them to analyse future scenarios.

8. Generation and attraction: converting territory into mobility

The first stage seeks to estimate how many trips each zone generates and attracts. Mobility does not appear spontaneously; it is closely related to the structure of the territory. Population, employment, economic activity, educational centres, hospitals, retail, industry, major facilities and land uses all generate different travel patterns.

A predominantly residential area generates certain types of trips, while a major employment centre attracts others, and a healthcare facility may have a completely different temporal distribution. The purpose of this stage is not merely to reproduce existing conditions, but to establish sufficiently robust relationships to allow future scenarios to be constructed.



If a zone gains 10,000 new inhabitants, if a new hospital is developed, or if a major area of economic activity is created, the model must be capable of translating those territorial changes into new mobility demand. In this way, urban planning and transport planning cease to be treated as independent processes.

9. Distribution: knowing how many trips exist is not enough; we need to know where they go

A city may generate a certain volume of trips, but the structure of the transport system fundamentally depends on the relationships between origins and destinations. Two neighbourhoods with similar populations may exhibit completely different behaviours: one may be strongly linked to the city centre, another to an industrial area, and another to surrounding municipalities within the metropolitan area.

Spatial distribution transforms generated and attracted trips into specific relationships between zones. It is precisely at this stage that large-scale mobility datasets offer some of their greatest potential, because many relationships that traditionally had to be inferred from relatively small samples can now be observed from much larger volumes of trips.



However, observing the present is not the same as explaining the future. Modelling remains necessary to understand how these relationships may change when land uses, infrastructure, travel times or the transport supply itself are modified.

10. Mode choice: probably one of the most important stages

Once we know who is travelling and where they are going, one essential question remains: how do they make that journey? Private car, public transport, bicycle, walking and other alternatives compete with one another depending on the specific conditions of each trip.

Mode choice depends on numerous variables: in-vehicle time, access time, waiting time, transfers, cost, frequency, reliability, parking availability, accessibility and even characteristics of the individual traveller. The relevant time is not simply time spent on board the vehicle, but the sum of access time to the stop or station, waiting time, travel time and access time from the final stop or station to the destination. The generalised cost equation for the complete journey must consider not only these different time components —which may themselves be weighted using specific coefficients— but also less tangible factors such as comfort and convenience, including price, frequency, safety, comfort, accessibility and other aspects.

This stage becomes particularly important when public policy seeks to increase public transport use. It is not sufficient to understand those who already use buses, metro or rail. We also need to identify which trips currently made by other modes could potentially transfer to public transport, and under what conditions.



This potential for modal shift is a critical variable because it will subsequently influence both the level of service required and revenue —or the success of the system in the case of fare-free public transport— and therefore the economic balance of the system.

11. Assignment: converting demand into actual use of the networks

The final stage distributes trips across the different networks. In public transport, it makes it possible to estimate routes used, transfers, passenger loads by section, main corridors and demand by service. In private traffic, it allows us to analyse how vehicles distribute themselves among the available routes.

The interaction between the two systems is particularly important in urban environments, where buses share a significant proportion of road space with general traffic. A route may have an excellent theoretical frequency and still provide a relatively uncompetitive service if it operates through corridors that are systematically affected by congestion.



For this reason, public transport planning and traffic engineering should not be understood as completely independent problems. The road network itself forms part of the operating conditions of public transport, and its performance ultimately affects both the quality offered to passengers and the cost of producing the service.

12. Before using a model to discuss the future, it must demonstrate that it understands the present

A model may contain thousands of links, hundreds of zones and visually impressive outputs without necessarily representing reality correctly. Before using it to assess future alternatives, it must be calibrated and validated using observed information.

On the road network, simulated flows can be compared with actual traffic counts, while public transport passenger loads can be compared with ticketing data or specific boarding and alighting surveys. Journey times and speeds can be validated using operational or GPS data, and occupancy levels, mode shares, corridor usage and temporal patterns can also be examined.

The greater the number of independent sources that can be used to validate the model, the greater the confidence that can be placed in its results. Calibration should not simply be an exercise aimed at achieving a particular percentage of statistical fit. It should be a process intended to determine whether the internal logic of the model adequately explains the observed operation of the system.



Only once it has reasonably passed that test does it make sense to use the model to formulate hypotheses about the future.

13. A strategic model should not be used to justify a decision that has already been made

The purpose of a model should not be to demonstrate that a particular alternative works, but to allow different alternatives to be compared before a decision is made. Once developed and validated, it can be used to assess route restructuring, new services, frequency modifications, interchanges, urban developments, restrictions on private vehicles, park-and-ride facilities, priority corridors or fare changes.

Its real value lies in allowing us to experiment and, if necessary, make mistakes within the model before making mistakes in the city. This capacity for comparison is precisely what turns modelling into a decision-support tool rather than merely a representation tool.



However, for that comparison to be complete, one further dimension must be incorporated that is still too often treated separately: the cost of each alternative and the resources required to make it viable.

14. From passengers to kilometres, hours, vehicles and staff

A strategic model may estimate that a particular route will carry 3,000, 10,000 or 20,000 passengers per day, but that result still does not tell us what resources will be required to provide the service. Estimated demand must be converted into actual transport production.

This requires the definition, among other variables, of route length, frequency, vehicle capacity, cycle time, number of trips, commercial and dead-running kilometres, number of vehicles required, driving hours and maintenance requirements. At this point, the mobility model connects directly with the operational model.

The same level of demand may be accommodated through different combinations of frequency and capacity. Higher-capacity vehicles may reduce the number of trips required, while higher frequency reduces waiting times and increases the attractiveness of the system, but also increases kilometres, operating hours and fleet requirements.



There is therefore not always a single technical solution. There is a range of possible solutions that must be assessed simultaneously from the perspective of the passenger, operations and the resources required to produce the service.

15. Economics should not enter the process only once the design has been completed

A relatively common practice is to design the network first and calculate its cost afterwards. I believe the process should be more iterative: the technical design and the economic and financial model should evolve together, because practically every operational decision has an economic consequence.

In simplified terms, the total cost of a service can be interpreted as the sum of fixed costs, costs associated with vehicle-kilometres, costs related to vehicle-hours, fleet and infrastructure costs, and other operating costs. This formulation helps to illustrate a fundamental issue: cost does not depend solely on the distance travelled, but also on the time required to produce that distance.

A route covering 20 kilometres in 60 minutes does not necessarily have the same operating cost as another route capable of covering the same distance in 45 minutes, even though the kilometres operated are identical. This is why variables initially regarded as traffic-engineering issues, such as congestion or signal priority, have a direct economic impact on public transport operations.



The economic and financial dimension should therefore not be used merely at the end of the process to check whether the proposed solution can be afforded. It should help determine from the outset which solution is best.

16. Commercial speed is also an economic variable

Suppose a route has a complete cycle time of 60 minutes and, through different priority measures, we manage to reduce it to 54 minutes. The cycle time has been reduced by 10%, and from the passenger's perspective this may mean a shorter journey time, improved reliability and potentially better regularity.

But the consequences do not end there. From an operational perspective, the route requires less time to complete each cycle and, depending on the timetable structure, this reduction may make it possible to maintain the same frequency with fewer resources, increase frequency with a similar fleet, or reduce the total number of vehicle-hours required.

Signal priority therefore ceases to be merely a traffic-engineering measure and becomes a measure of operational and economic efficiency. The same applies to reserved lanes, the elimination of bottlenecks, parking reorganisation, stop relocation and certain modifications to intersections.



Commercial speed should therefore be analysed simultaneously as a service-quality indicator, a capacity variable and an economic operating variable.

17. Microsimulation makes it possible to quantify this relationship

It is precisely at this point that microsimulation becomes particularly useful. While macrosimulation allows the city and its main corridors to be studied, microsimulation enables much more detailed analysis of intersections, traffic signal control, queues, reserved lanes and the specific interaction between public transport and general traffic.

Its results, however, do not need to end with a table of delays. Suppose an intervention reduces the journey time of a route by 90 seconds and 200 services pass through that location every day. The accumulated reduction is equivalent to 18,000 seconds per day, or 5 operating hours every day.

When this effect is multiplied across the year, what initially appears to be a relatively minor improvement at a single intersection can become operationally and economically significant. Microsimulation therefore allows seconds or minutes saved to be converted into vehicle-hours and subsequently transferred into the operational model and the economic analysis.



This connection between different modelling scales is particularly useful because it allows interventions to be assessed not only in terms of their effect on traffic, but also in terms of their contribution to public transport productivity.

18. Fares cannot be analysed independently from demand either

Revenue constitutes the other fundamental part of the system. In simplified terms, fare revenue can be expressed as the product of the number of passengers and the average revenue per trip, but neither of these variables necessarily remains constant.

Changing the fare structure can affect average revenue and simultaneously modify demand. A reduction in fares may increase passenger numbers, although not necessarily by enough to compensate for the reduction in revenue per passenger; a fare increase may produce the opposite effect.

Furthermore, modern systems generally include different ticket products, travel passes, discounts, transfers, integrated fares and social fares. The economic and financial analysis should therefore be connected to the demand model itself, since a change in the generalised price of a journey may alter mode choice and, consequently, the number of public transport users.



Fares are not merely a financial decision. They are also a mobility policy tool.

19. Fleet is an operational, technological and financial decision

Vehicle selection and fleet renewal provide another good example of the interdependence between the different dimensions of the system. When selecting a particular fleet, we are simultaneously making decisions about capacity, accessibility, emissions, energy consumption, maintenance, availability, useful life, required facilities and investment.

Comparing technologies exclusively on the basis of purchase price is therefore insufficient. One alternative may require a higher initial investment but offer lower energy costs; another may require new infrastructure or have a different maintenance structure. The analysis should therefore consider the entire life cycle of the solution.



Furthermore, the number of vehicles required is not an independent variable. It depends on frequencies, cycle times, reserve requirements and the operational structure of the network, once again directly connecting infrastructure and traffic decisions with investment requirements.

20. The cost of a new route must be assessed together with the mobility it generates

Suppose our model identifies a relationship that is currently poorly served and estimates that a new route could attract 2,000 passengers per day. From a mobility perspective, that result may initially appear favourable, but many questions still need to be answered.

How many annual kilometres will the service require? How many trips will need to be operated? How many vehicles and driving hours will it involve? Will additional investment be required? What revenue is expected, and what level of public contribution will be required to support the service?

And, above all, there is a particularly interesting question: could we capture a similar proportion of that demand by reorganising the existing network at a lower marginal cost?



This is where planning and economics genuinely begin to function as a single optimisation problem. The objective should not be simply to maximise passenger numbers or exclusively to minimise cost, but to find an appropriate balance between demand, territorial coverage, journey time, frequency, quality, required resources and economic sustainability.

21. The economic model must also assess scenarios and risks

Public transport services are provided over long periods, and the future will never correspond exactly to our forecasts. Demand may change, energy or labour costs may increase, new urban developments may appear, or significant service modifications may become necessary.

An economic and financial model should therefore not produce only a central scenario. It should allow sensitivity analyses to examine what happens, for example, if demand is 10% lower than forecast, energy prices increase, the number of kilometres operated rises, or commercial speed falls.

These analyses help identify the variables that genuinely determine the viability of the system and assess how resilient a particular solution is to reasonable changes in the initial assumptions. A technically excellent network that is economically extremely sensitive may become problematic only a few years after implementation.



Sustainability should therefore not be assessed only under the scenario we consider most likely, but across a reasonable range around that scenario.

22. The financing structure can also influence how the system operates

Where public funding is involved, the method used to calculate it is not simply an administrative matter, because it can create different incentives for the operator. A model based exclusively on production may incentivise additional kilometres, while a model based only on demand may create difficulties for socially necessary but low-demand routes.

Funding mechanisms may incorporate variables related to vehicle-kilometres operated, trips completed, regularity, punctuality, availability, commercial speed, demand or perceived quality. There is no single correct structure, since the appropriate model will depend on the management arrangement and the allocation of risk between the parties.

The important point is to recognise that the economic mechanism may influence operational behaviour. Public funding can be merely a way of paying for the service, or it can also be designed as a tool for aligning the objectives of the operator with those of the public authority.



The financial architecture should therefore form part of the overall design of the system rather than being treated as an entirely separate element from technical planning.

23. Not every service has to maximise profitability

Integrating the economic dimension into planning should not lead to an excessively simplistic interpretation either. Public transport is a service of general interest, and some routes may present less favourable economic indicators while still fulfilling essential functions related to accessibility, territorial cohesion, social inclusion or access to basic services.

Indicators such as cost per passenger or farebox recovery should therefore always be interpreted within the objectives established for the system as a whole. A low-demand route may be entirely necessary and fully justified from a public policy perspective.

The difference is that we should know how much it costs, what function it performs and why we have decided to maintain it. Efficiency does not necessarily mean eliminating whatever has the highest unit cost. It means understanding what outcomes are obtained in return for the resources used and whether those same objectives could be achieved more efficiently.



Economic evaluation should help us make better decisions, not replace the social and territorial objectives that justify the existence of public transport itself.

24. Planning should become a continuous cycle

All models contain assumptions and, once an intervention has been implemented, we must therefore verify what actually happened. If a route is modified, we should analyse how demand responded; if signal priority is introduced, we should measure the actual improvement in commercial speed; if frequency is increased, we should verify whether new passengers were attracted; and if fares are modified, we should observe simultaneously how both demand and revenue evolved.

The same applies to investment decisions. If a new fleet is acquired, actual operating and maintenance costs should be compared with those originally assumed in the model, and any significant deviations should be analysed.

Planning should not follow a linear sequence of data → study → intervention → end. Instead, it should function as a continuous cycle of observe → model → design → evaluate economically → implement → measure → recalibrate.



This final stage is fundamental because every implemented decision generates new information. If that information is fed back into the models, the actual operation of the system itself can progressively improve the quality of future decisions.

25. From studying individual routes to managing systems

The evolution of public transport engineering is not simply about using new software tools. Above all, it represents a change in approach, in which different sources of information and different levels of analysis cease to operate as independent compartments.

Surveys continue to provide valuable information, but they can now be complemented by millions of observations from other sources. Mobile phone data help us understand the territorial structure of mobility; ticketing allows us to analyse actual system use; operational systems and GPS provide insight into operations; boarding and alighting counts allow passenger loads to be validated; and traffic counts describe the road traffic with which public transport shares space.

Origin-destination matrices structure that demand; macrosimulation allows scenarios to be assessed; microsimulation enables detailed analysis of specific interventions; the operational model converts passengers into kilometres, hours, vehicles and staff; and the economic and financial model converts those resources into costs, investment, revenue and funding requirements.



All these tools observe the same system from different perspectives. Their real potential emerges when we stop using them as separate studies and begin connecting them within a single planning process.

26. Conclusion: planning mobility, operations and economics as a single system

Perhaps the most important question in public transport planning is no longer simply how we can improve a particular route, or even how we can improve an existing network. The question should be framed more broadly: how can we serve the mobility needs of a city more effectively while using the available resources efficiently and sustainably?

Answering this question correctly requires us to understand demand, operations, traffic, infrastructure and economics simultaneously, because in public transport almost no decision produces only one effect. A priority measure may reduce cycle time; that reduction may decrease resource requirements or allow frequency to be increased; a higher frequency may reduce waiting time and improve the competitiveness of the service; and increased competitiveness may modify demand, revenue and future service requirements.

The same applies in the opposite direction. A fare change may alter demand; that change may modify passenger loads on particular routes and require more or less capacity; that new level of service will affect the kilometres and hours operated and ultimately modify the costs of the system once again.

That is, in reality, the system we are planning. Not isolated routes, not simply buses, and not simply costs, but a mobility, production and financing system whose components are permanently interconnected.

Understanding those relationships, having sufficient data to characterise them, and being able to model their consequences before intervening is probably one of the most interesting and complex tasks in public transport engineering.



Lluis Sanvicens, 2026

Photographs taken in Warsaw in 2026. Source: own archive.