As Product Manager of Global Street Data Products at Pitney Bowes, one of the questions I get asked on a regular basis is: "I know this road, at rush hour it must get tens of thousands of cars on it - yet your StreetPro Traffic Data only shows 1200 cars on this section between 9am and 10am on weekdays. How come your data differs so much from reality?" Let's shine a light on the two kinds of Traffic data commonly used at Pitney Bowes and what they can offer.
Traffic Census data is the traditional method of capturing traffic data, where a person sits at a junction or intersection of a road and manually records the number and types of vehicles that pass by at different times of the day and days of the week. This method provides a very detailed view of the traffic levels at a particular point in time in a particular location. Combine this with data from automated traffic counters on the surrounding network and a fairly detailed estimate of the traffic levels and speeds in a location can be obtained.
Floating Car Data is a more modern method for estimating traffic data, and consists of collecting positioning device data (i.e. GPS data) from within vehicles. These devices provide speed, time and direction of travel data, and can also be counted to provide an indicator of relative traffic levels at any one time on the network. Satellite Navigation devices, mobile phones, in-car navigation devices, telematics sensors all contribute data to the floating car dataset.
The limits of Traffic Census Data
Collecting traffic census data is an expensive way to build traffic data, as it requires that people sit at junctions or street intersections to collect the data, or manually place and retrieve automated traffic counters across the roads. Consequently, most traffic census data are derived by combining the actual counts with data obtained from the automated traffic counters to estimate an average day's traffic count. In countries that do this kind of surveying, the focus is very much on the major road network. Minor roads significantly outnumber the major roads, and it is simply unfeasible and cost prohibitive to sample these roads extensively. Agencies carrying out census work will therefore take a representative sample and use models to derive counts on these roads where data collection is not possible.
The limits of Floating Car Data
Floating car data has cost benefits because government transport agencies don't need to purchase the devices that provide the data. Instead, they merely collect the data from vehicles and process them. These data feeds provide near real-time and historic views of the congestion on the entire road network. This is how live traffic services used in our navigation devices are able to tell us that there is congestion ahead and we should take an alternative route. Now consider the scale required to cover the whole road network. Billions of data points from devices (probes) daily are required to provide accurate estimates of traffic flow along all streets within the street network. Our StreetPro traffic product is built from data sourced from over 500 million devices supplying over 11 billion measurements every day worldwide. That's a lot of data, but it will never give you an accurate picture of the actual traffic levels for one simple reason. Not every vehicle on the road has a positioning device in it! The pool of available data is also reduced by the impact of human behaviour. How often do you switch on your satellite navigation device to go to the local shops or on the commute to work? How often do you plug your satellite navigation device into the internet and update it? How often do you take your car to the garage for them to plug it in? These kinds of delayed data feeds mean that floating car data needs to be regularly updated to take advantage of the lag from human behaviour.
If Traffic Census data is based on the ground truth, why use Floating Car data at all?
As with many things to do with data it all comes down to what you want to understand from your data, and whether or not the data you need for your analysis are available. Do you want to know the average number of vehicles that travel down a road? Or does your business need to be located at a site where X number of vehicles pass by each day? The census data will be your best option here, and can help you plan around the sheer volume of traffic. Can traffic census data pick up changes that happen over certain periods of time? Not unless people are employed to manually count the traffic at the location of interest.
Alternatively, floating car data can help understand the relative traffic levels at scale and over time (hourly, monthly). Do you want to understand the seasonal variations in relative traffic levels in a location? Do you want to compare different locations and make decisions based on the relative congestion levels around them? Do you want to compare the levels of congestion on a road at different times of the day? Do you want to assess the impact of short to medium term change on the relative congestion levels in the area? Floating car data can help to answer these questions.
------------------------------
Daniel Edwards
Product Manager, Location Intelligence Data
Pitney Bowes Software Ltd
HENLEY-ON-THAMES
------------------------------