A traffic model is a tool for analyzing the major ways people travel using a computer program. There are a variety of software packages available, but generally they are all based on the same three parts. These are:

1. A representation of the road network
2. Socio-economic data
3. A mathematical formula to distribute trips

The network represents the major routes for getting around within the study area. In simple models this is the road system. Overall, the network shows the supply of transportation.

The road network is represented by nodes (intersections) and links which connect them (roads). Each node and link also have information about their real physical data, such as capacity and speed. The network ends at external stations, which represent connections to the "outside world".

Land use in the area being studied is a major factor in how many trips people are likely to make. Rates have been determined for roughly how many trips will leave different general types of land uses, such as shopping strips or residential neighborhoods. The model area is split up into areas of similar land use, called traffic analysis zones (TAZs). These zones are represented in the network by special nodes called centroids.

The demand for transportation is also based on socio-economic information about each TAZ. Examples of the types of information used in a traffic model are the number and type of jobs, and the number of households.

A mathematical equation uses the network and land use information to simulate traffic in the region. This equation determines the number of trips which will be made and the route each will take. This is only for automobile trips in most models. The process builds on two basic ideas, which are: Job sites, shops, and homes produce and attract trips. Traffic analysis zones with a lot of any of these naturally get a lot of trips. Also, people are willing to travel further to get to work than to pick up a pizza. The amount and type of land uses in a zone determines how "attractive" it is to trips. This basic concept gives the equation its name: the gravity model. Drivers will usually take the shortest, fastest routes. As traffic congestion makes one route less desirable, other routes are used.

Since a traffic model is just a simulation, it is essential to check that it is realistic. The process of testing and adjusting the model is called calibration. An important part of this is comparing the results to known data, such as real traffic counts. This is done for a specific base year for which data is available, such as 1997.

The results from a traffic model include the number of cars on any given road and where roads are congested with traffic. This provides some standard transportation measures for an area. These include Vehicle Miles Traveled (VMT) and Vehicle Hours of Delay (VHD). A further step can translate these results into tailpipe emissions for measuring impacts on air quality.

A model can be run with changes to its land use information and/or its road network. Using forecasts of socio-economic data produces information about future traffic. Using forecasts and no changes to the traffic network (other than those already approved) provides the no-build scenario.

Traffic models are also used to test ideas for changes in the road network. This helps analyze how a change in one road might change traffic there and in the broader area. The results of running different alternatives can be compared as an aid in deciding which works best.

It is essential to keep in mind that a traffic model is only a tool, and an imperfect one at that. People need to consider many factors in making recommendations on transportation decisions. Many of these factors are either not adequately or not at all included in traffic models.