Transportation models support billions of dollars in infrastructure decisions every year — yet they are routinely blamed when projects produce unexpected outcomes. The problem is rarely the software. It's how models are developed, interpreted, communicated, and applied.
Models Are Decision Support Tools, Not Crystal Balls
What Models Actually Do
A transportation model does not predict the future. It estimates what is likely to happen under a defined set of assumptions. Change the assumptions — change the forecast. That is not a flaw. That is the entire point.
Models reduce uncertainty. They never eliminate it.
Inputs That Drive Every Forecast
- Population and employment projections
- Land use and zoning assumptions
- Network geometry and roadway supply
- Transit service levels and coverage
- Travel costs and fuel pricing
- Travel behaviour and mode preferences
- Active policies and demand management
❌ Wrong Question
What will happen?
✅ Right Question
What is likely to happen if we implement this project under these assumptions?
Poor Inputs Create Poor Outputs
No modelling software — however sophisticated — can overcome inaccurate or outdated assumptions. Garbage in, garbage out remains the most fundamental truth in transportation modelling. When inputs are flawed, every downstream output inherits that flaw, often invisibly.
Typical Input Problems
- Outdated or unrepresentative traffic counts
- Incorrect or optimistic zoning forecasts
- Unrealistic development timing assumptions
- Incorrectly coded transit service
- Missing or incomplete roadway network changes
- Poor origin-destination survey data
Questions to Ask Before Modelling
- Are traffic counts current and representative of typical conditions?
- Are development forecasts realistic and phased correctly?
- Is transit service coded accurately for the horizon year?
- Are demographic projections sourced from authoritative datasets?
Poor Inputs
Outdated counts, flawed forecasts, missing network data
Model Processing
Sophisticated software amplifies and propagates every flaw
Unreliable Outputs
Flawed forecasts, misguided decisions, wasted investment
Poor inputs do not stay contained — they contaminate every analysis, every scenario, and every recommendation that flows downstream from the model.
Asking the Model Questions It Was Never Built to Answer
Every transportation model is designed to answer a specific set of planning questions. Using a model beyond its intended scope does not produce better answers — it produces misleading ones. Understanding a model's boundaries is just as important as understanding its capabilities.
Toll Revenue Estimation
Credible toll revenue forecasts require explicit representation of Value of Time (VoT), toll choice behaviour, and route choice sensitivity. Without these, revenue estimates are speculative at best.
Ride-Hailing Demand
If Uber, Lyft, and similar services are not represented in the mode choice structure, the model cannot estimate mode shifts, ride-hailing demand levels, or induced trip generation from these services.
Micro-Mobility & Emerging Modes
E-scooters, bike-share, and autonomous vehicles require specific behavioural parameters. Models built before these modes existed cannot reliably forecast their impacts without explicit recalibration.
A thermometer isn't broken because it can't measure wind speed. Neither is a transportation model. Know what your model measures — and what it doesn't.
Skipping Calibration and Validation
Software defaults rarely represent local travel behaviour. Every city has unique driving patterns, transit usage levels, freight dynamics, and congestion characteristics. A model running on default parameters is an unproven model — regardless of how sophisticated the underlying software may be.
Calibration
During calibration, the model's behavioural mechanisms are adjusted — trip generation rates, gravity model decay functions, mode choice utilities, and assignment parameters — until the model reproduces observed base-year conditions. The focus is on tuning the mechanisms that drive travel behaviour.
Validation
During validation, the resulting model outputs are compared against independent observations not used in calibration, and checked for reasonableness and credibility. This confirms that the model's outputs — volumes, travel times, mode shares — are accurate and defensible.
Treating Model Outputs as Absolute Truth
False Precision in Practice
Models routinely report outputs like:
- 18,742 vehicles/day
- 14.3 minutes of travel time
- 97.2 seconds of intersection delay
These numbers look authoritative. They are not exact. They are estimates — and presenting them as certainties misleads decision-makers.
Sources of Forecast Uncertainty
- Population and employment growth rates
- Economic conditions and business cycles
- Fuel prices and energy transitions
- Transit funding levels and service changes
- Behavioural shifts (remote work, e-commerce)
- Emerging technology adoption curves
- Land development timing and phasing
Scenario Ranges
Present low, base, and high forecasts that bound the likely range of outcomes rather than a single deterministic result.
Sensitivity Analysis
Test how forecasts change when key assumptions shift — population, mode share, fuel cost — to reveal which inputs most influence outcomes.
Transparent Assumptions
Document and communicate every major assumption explicitly so decision-makers understand the basis — and the limits — of each forecast.
Asking the Wrong Planning Question
A transportation model cannot compensate for poor problem definition. When the wrong question is asked at the start, even the most rigorous modelling effort will produce answers that are technically correct but strategically irrelevant. The framing of the question determines the value of the analysis.
❌ A Weak Question
Should we widen the road?
This question presupposes a solution before the problem has been defined. It constrains the analysis and limits the alternatives considered.
✅ Stronger Planning Questions
- What specific problem are we trying to solve — delay, reliability, safety, or accessibility?
- Who is affected, and how does this vary by mode or user group?
- What alternatives could address this problem before adding new lanes?
- Does bus priority, signal optimization, or demand management achieve comparable outcomes at lower cost?
Ignoring Professional Judgment
Transportation models are analytical tools. They are not decision-makers. The most common — and most costly — mistake in transportation planning is allowing a model output to substitute for experienced professional judgment. Results must always be interrogated, not just accepted.
Transportation Models
Engineering Judgment
Field Observations
Local Knowledge
Stakeholder Input
Operational Experience
Experienced practitioners always ask: Does this result make sense? Does it match what we observe in the field? Is there a coding issue, a missing behaviour, or a policy shift that could fundamentally alter travel patterns? These questions are not second-guessing the model — they are essential quality control.
What Great Transportation Modelling Looks Like
Successful transportation projects follow a disciplined, transparent process that integrates rigorous modelling with clear communication and professional judgment. Great modelling builds confidence in decisions — not just in forecasts.
Define the Decision
Clearly articulate the infrastructure or policy decision that modelling must inform.
Ask the Right Question
Frame the planning question around the actual problem — not a presumed solution.
Build Quality Datasets
Collect current, representative traffic counts, demographics, land use, and transit data.
Select the Appropriate Model
Match the model type and scope to the specific planning questions being evaluated.
Calibrate and Validate
Adjust parameters to reproduce observed conditions and verify model reasonableness.
Test Multiple Scenarios
Evaluate low, base, and high cases alongside alternative investment strategies.
Communicate Assumptions and Uncertainty
Present ranges, sensitivities, and key assumptions transparently to all stakeholders.
Apply Engineering Judgment
Combine model outputs with field observations, local knowledge, and professional experience.
Make Evidence-Based Decisions
Use the full analytical picture to support defensible, transparent infrastructure investments.
Better Models Lead to Better Decisions
❌ How Models Fail Decision-Makers
- Treated as black boxes that require no scrutiny
- Built on weak, outdated, or unrepresentative assumptions
- Applied to questions they were never designed to answer
- Presented without ranges, uncertainty, or sensitivity analysis
- Interpreted without professional engineering judgment
✅ How Successful Organizations Use Models
- Invest in quality data collection and validation
- Calibrate and validate rigorously before evaluating alternatives
- Use scenario-based analysis to bound uncertainty
- Communicate assumptions and limitations transparently
- Combine modelling with engineering expertise and field observations
"The goal of transportation modelling is not to predict the future with certainty — it is to improve today's decisions by making future outcomes more understandable."
The value of transportation modelling lies not in producing precise numbers, but in helping decision-makers compare alternatives, understand trade-offs, and make more informed, defensible investments.
Why Transportation Models Fail Decision Makers — And How to Avoid It