Blog

Why Transportation Models Fail Decision-Makers (And How to Avoid It)

‍Transportation models have become an essential part of planning and infrastructure investment. They help forecast future travel demand, evaluate transit improvements, assess development impacts, and compare alternative scenarios.

Yet despite increasingly sophisticated software and larger datasets, many projects still end with frustrated stakeholders asking questions like:

• “Why didn’t the model predict this?”

• “Can we trust these results?”

• “The model says one thing, but our experience says another.”

The problem usually isn’t the model. ‍

The problem is how the model is developed, interpreted, and used during decision-making.

A Transportation Model Is a Decision Support Tool—Not a Crystal Ball

One of the biggest misconceptions in transportation planning is that models predict the future. ‍

They don’t.

‍A transportation model is simply a structured representation of how people and vehicles are expected to travel under a defined set of assumptions. Those assumptions include land use, population growth, transit service, roadway networks, travel behaviour, and many other variables.

Change the assumptions, and the results change.

Good modelling isn’t about producing a single “correct” forecast. It’s about helping decision-makers understand the likely outcomes under different scenarios and the uncertainty surrounding them.

The most valuable question is rarely:

“What will happen?”

‍Instead, it is:

“What is likely to happen if we make this decision?”

Where Transportation Models Go Wrong

1. Poor Quality Inputs

Every model depends on the quality of its input data.

If traffic counts are outdated, land-use assumptions are unrealistic, or development forecasts are inaccurate, even the most advanced modelling software will produce misleading outputs.

This is the classic “garbage in, garbage out” problem.

Before running any model, it’s worth asking:

• Are the traffic counts representative?

• Are future developments realistically represented?

• Has transit service been coded correctly?

• Are demographic forecasts current?

Reliable outputs begin with reliable inputs.

2. Expecting the Model to Answer Questions It Was Never Built to Answer

One of the biggest disconnects between project teams and decision-makers is assuming that a transportation model can answer any transportation question.

In reality, every model has a defined scope. It is built to represent certain travel behaviours, choices, and network characteristics—and anything outside that scope cannot be estimated reliably.

For example:

• A client may ask, “How much toll revenue will this project generate?” But if the travel demand model does not include a Value of Time (VoT) distribution or a behavioural model that represents travellers’ willingness to pay tolls, the model cannot credibly estimate toll diversion or revenue. It may predict traffic volumes, but not the behavioural response required for a tolling analysis.

• Another common request is to evaluate ride-hailing services such as Uber or Lyft. If the model was developed before ride-hailing was represented as a travel mode—or if it groups those trips under private auto—it simply has no mechanism for estimating shifts to or from ride-hailing.

This is not a limitation of the software. It is a limitation of the model specification.

Every transportation model is designed to answer a particular set of planning questions. Asking it to answer questions beyond that design is like asking a thermometer to measure wind speed. The instrument isn’t failing—it simply isn’t measuring the phenomenon you’re interested in.

3. Skipping Calibration and Validation

A common mistake is assuming that software defaults accurately represent local conditions.

They rarely do.

Every city has unique travel patterns, driver behaviour, transit usage, freight activity, and network characteristics. That’s why calibration and validation are essential parts of the modelling process.

Calibration adjusts model parameters so the model reproduces observed conditions, while validation tests whether the model responds reasonably to changes and can produce credible forecasts. Transportation agencies such as the Federal Highway Administration emphasize that calibration should always occur before alternatives are evaluated because default parameters alone are rarely sufficient to represent local conditions. (FHWA Operations)

Without proper calibration, the model may appear sophisticated but have little predictive value for the decisions it is intended to support.

4. Treating Model Outputs as Absolute Truth

Transportation models often produce outputs with impressive precision:

• 18,742 daily vehicles

• 14.3-minute travel time

• 97.2-second delay

These numbers can create a false sense of certainty.

In reality, every forecast contains uncertainty.

Population growth may differ from projections.

Travel behaviour changes.

Economic conditions shift.

New developments occur earlier—or later—than expected.

Good analysts communicate ranges, assumptions, and confidence, rather than presenting a single forecast as an unquestionable answer.

5. Asking the Model the Wrong Question

Sometimes the model is technically correct, but the project team asks the wrong question.

Instead of asking:

“Should we widen this road?”

A better question might be:

• What problem are we trying to solve?

• Is delay the biggest issue?

• Could transit priority achieve similar benefits?

• Would active transportation investments change travel demand?

• Are there operational improvements before major capital investments?

A model cannot compensate for poor project framing.

The quality of the decision depends on asking the right questions before any analysis begins.

5. Ignoring Professional Judgment

Transportation models are powerful analytical tools.

They are not replacements for engineering judgment.

‍Experienced planners and engineers recognize when results deserve further investigation.

Does a forecast seem inconsistent with observed travel patterns?

Does a scenario produce unexpected network effects?

Are there policy or behavioural changes the model cannot fully capture?

Good practitioners use modelling alongside field observations, stakeholder input, operational knowledge, and engineering experience.

The strongest recommendations come from combining all of these perspectives.

The Best Models Build Confidence—Not Just Numbers

‍Ultimately, transportation models should reduce uncertainty—not create more of it.

A successful modelling project doesn’t simply generate forecasts.

It gives planners, engineers, elected officials, and stakeholders confidence that major infrastructure decisions are based on sound evidence, realistic assumptions, and transparent analysis.

When modelling is approached as part of a broader decision-making process—not as an isolated technical exercise—it becomes one of the most valuable tools available for transportation planning.

Final Thoughts

‍Transportation models don’t fail because the software is inadequate.

They fail when they are treated as black boxes, built on weak assumptions, or presented without sufficient context.

The organizations that consistently make better transportation decisions are those that view modelling as one component of a larger analytical process—one that combines quality data, rigorous calibration, scenario testing, clear communication, and experienced professional judgment.

Because in the end, the goal isn’t simply to build a better model.

It’s to make better decisions.