Building Demand Models for Practical Transit Planning

What demand modeling helps agencies understand

In practice, demand modeling helps agencies connect service changes to expected rider response.

• How many riders may use a proposed service.
• Which origins and destinations are better connected.
• Where improved service may unlock latent demand.
• How travel times and access change for different communities.
• How a proposed scenario compares to the current network.

The same modeling framework is designed to work across a wide range of planning environments: agencies with limited local datasets, agencies with detailed ridership and operations data, and agencies that already have their own regional demand models.

This post goes deeper on one part of the Planning Platform. For the full system context, read the Planning Platform overview.

Starting with a general activity-based mode choice model

At the foundation of the platform is a general activity-based mode choice model.

The model is designed to estimate how people choose between available travel options based on the trips they need to make, the places they need to reach, and the service available to them. It considers the underlying travel market:

• Where people live.
• Where they work.
• Where they go to school.
• Where they access healthcare and daily needs.
• How different transportation options compare for those trips.

Why latent demand matters

This is important because transit demand is not just a reflection of the current network. Existing ridership tells us who is using today's service, but it does not fully capture latent demand: people who might use transit if the service were:

• Faster.
• More direct.
• More frequent.
• More reliable.
• Better connected to the places they need to go.

A planning tool that only looks at existing ridership can miss those opportunities. A demand model should help agencies understand not only where transit is working today, but where better service could unlock new ridership, improve access, or shift trips from other modes.

That is the role of the base model. It gives the platform a regional understanding of travel behavior that can support scenario testing even before every local dataset has been added.

Tuning the model with local agency data

The base model becomes more powerful when it is tuned with local data.

Every agency has a different data environment. Some agencies may have detailed stop-level boardings and alightings by time of day. Others may have:

• Route-level averages.
• Monthly summaries.
• APC samples.
• Survey results.
• A smaller set of planning spreadsheets.
• Third-party mobility datasets.
• TNC data.
• A separately maintained regional demand model.

The platform is designed to use those inputs when they are available.

Calibrating to local conditions

That calibration process helps align the model with local conditions:

• The actual ridership patterns on the network.
• The places where service is over- or under-performing.
• The times of day when demand is strongest.
• The markets where improved service may create the biggest response.

The important point is flexibility. The model should not require every agency to have the same data before planning can begin. It should provide a strong starting point, then improve as more local information is available.

Supporting many levels of ridership detail

Ridership data can arrive in many forms.

For some agencies, the available data may be coarse: total route ridership, average weekday ridership, boardings by service period, or periodic counts from manual surveys. For others, the data may be much more granular: stop-level boardings and alightings, APC records, ridership by hour, ridership by minute, trip-level loads, or exact observed on-offs.

The system is designed to work across that range.

Using coarse and detailed ridership data

Coarser ridership data can help calibrate overall demand levels, route productivity, and broad patterns of use. More granular ridership data can help tune stop-level demand, time-of-day variation, directional loads, transfer behavior, and the relationship between service quality and rider response.

Both are useful. The question is not whether an agency has a perfect dataset. The question is how to use the data it does have in a way that improves the forecast and supports better planning decisions.

By accepting ridership inputs at different levels of detail, the platform can support agencies that are just beginning to formalize their data workflows as well as agencies with highly detailed operations and passenger-counting systems.

Incorporating broader mobility datasets

Transit ridership is only one view of regional travel demand.

To understand where better service may be needed, agencies may also want to look at broader travel markets: trips currently made by car, taxi, TNC, walking, biking, paratransit, shuttle, or other modes. These datasets can help identify places where transit demand may exist even if current ridership is limited.

Ontra can ingest a range of mobility inputs, including TNC data and mobile location-derived datasets such as Replica. These inputs can help the model understand origin-destination patterns, trip purposes, time-of-day demand, and travel markets that are not fully visible from transit boardings alone.

Planning new or redesigned service

That is especially important for planning new or redesigned service. If an agency is evaluating a new microtransit zone, a redesigned fixed-route corridor, a shuttle connection, or a change to paratransit support, it needs to understand the trips people are already trying to make across the region—not just the trips currently captured by the existing transit network.

Broader mobility datasets help connect service planning to the actual movement patterns in a community.

Bringing your own demand model

Some agencies and regions already have a demand model they trust.

Ontra can support that as well. Agencies can bring their own demand model as long as it is regional and includes latent demand, not just observed transit ridership.

That distinction matters. A model that only reflects current transit trips may be useful for validating existing service, but it is less useful for evaluating new networks. To support service recommendations, the model needs to understand the larger market of trips that could shift to transit under different service designs.

When an agency brings its own model, Ontra can use that model as part of the planning workflow rather than forcing the agency to start over. The goal is to make existing investments in regional modeling more useful for day-to-day service planning, scenario testing, and recommendation generation.

From demand forecasts to service recommendations

Demand modeling is not an isolated analysis step. It is what allows the rest of the platform to make better recommendations.

When the platform tests a route change, stop pattern, service zone, or frequency adjustment, the demand model estimates how riders may respond. It can help answer questions like:

• How many riders may use this service?
• Which origins and destinations are better connected?
• Where might improved service unlock latent demand?
• How do travel times and access change for different communities?
• What tradeoffs exist between ridership, coverage, cost, and equity?
• How does a proposed scenario compare to the current network?

Those forecasts then feed into scenario comparison, optimization, and results analysis. The recommendation engine can search for service designs that perform well not just operationally, but in terms of how people are expected to use them.

That is what makes demand modeling central to the Planning Platform. It connects the supply side of planning—routes, stops, zones, vehicles, frequencies, costs—to the rider side of planning: where people need to go, what choices they have, and how the network can serve them better.

Making modeling usable in everyday planning

A sophisticated demand model is only valuable if planning teams can actually use it.

That means the model has to fit into the workflow. It has to:

• Work with real agency data.
• Support scenario iteration.
• Produce outputs people can inspect, compare, and explain.
• Help planners communicate tradeoffs, not just calculate forecasts.

That is why Ontra is building demand modeling as part of the Planning Platform rather than as a standalone technical exercise. Planners should be able to define a scenario, run a forecast, compare the results, adjust assumptions, and keep iterating without losing context.

Meeting agencies where they are

The model should also make planning more accessible across different agency contexts. Some teams may begin with public datasets and high-level ridership summaries. Others may bring detailed APC feeds, TNC data, mobile location-derived datasets, or a regional demand model. The platform is designed to meet agencies where they are, then improve the forecast as more data becomes available.

Better demand models, better planning decisions

Better demand modeling changes the planning process.

It helps agencies move beyond questions like, "How many people ride this route today?" and toward questions like:

• Where is demand underserved?
• Which changes would make transit more useful?
• How would riders respond if we redesigned the network?

Those are the questions that matter when agencies are trying to improve service, allocate limited resources, and make defensible decisions.

Clearer judgment, better recommendations

Demand modeling does not replace planner judgment. It gives planners a clearer view of the travel market they are planning for. It helps them:

• Evaluate both current riders and potential riders.
• Compare scenarios based on expected outcomes.
• Give the optimization engine the information it needs to surface recommendations that are grounded in how people actually move.

That is the future we are building toward: transit planning that is more predictive, more practical, and more connected to the real travel needs of the communities agencies serve.