leaderboard left
leaderboard right
content topleft content topright

Painting the Traffic Picture


Traffic is being drawn in more places every day. From Google Maps to your local TV news, from GPS devices, to iPhone apps– everyone wants traffic information. Now I love data, stats, figures, and charts as much as the next guy but trying to interpret traffic data is almost as frustrating as traffic itself. Here is my open call to the industry to clear the lane of confusion, my suggestion on how to depict traffic so that drivers can get meaningful and actionable data from the pretty traffic maps.

The Standard Traffic Map

Consider this map– a fairly standard depiction of “live” traffic in the Boston area from Google Maps.


Does anybody actually understand what those colors mean in a way that gives you actionable information? Sure, black/red = pretty much stopped. Red = slow. Orange = not perfect. Green = fast. That’s nearly what the legend tells us. But I still don’t know what that means.

To me, it appears that most of the highways are green. I’d expect them to be “fast” (green). And most of them are. Most of the smallest surface streets are red. Most of those roads have a 25mph speed limit and have lots of traffic lights so I’d expect them to be slow. In a glance, the colored traffic information doesn’t show me any more information than I could have deduced from the road classifications themselves. Big fat orange line = fast. Yellow line = not perfect. White = moving slow. Grey thin line = crawl.

In other words, I can make a pretty good estimation of how fast cars will be moving just by the road classification– I don’t need a “traffic” map to tell me that. You almost never see “green” paint on the smallest roads, yet they are very frequently in red. The more secondary roads that start to get covered by traffic data, the stranger this will look, especially to those familiar with the area.


I’m playing dumb a little bit here… buried in the Google Maps Help Docs it does say this.

  • Green: more than 50 miles per hour
  • Yellow: 25 – 50 miles per hour
  • Red: less than 25 miles per hour
  • Gray: no data currently available

This helps make my point above that most of the time this traffic data doesn’t tell us anything more than the road classification which we could have determined without the traffic map.

A Matter of Perspective

Not surprisingly, people look at traffic from different perspectives based upon where they live and were on the map they are looking. To someone who lives in Los Angeles, drawing a red line on the 405 at 5:00pm is simply confirming that a bear does in fact poop in the woods. There is no information there. To a tourist not familiar with the area that happens to be driving through, the red line is important.

So there are two very distinct categories of people who are seeking traffic data.

  1. Commuters, who know and understand what normal traffic volumes are. These people don’t really care how fast or slow traffic is moving– they want to know how it differs from normal. They want to know if an accident on their typical route will cause them an out of the ordinary delay.
  2. Travelers have no benchmarks for how traffic typically is. So they are more concerned with how fast traffic is moving (normal or not) so they can make route judgements.

Bryan Mistele of INRIX recently said this to GPS Business News.

Painting colors on a traffic map is easy to do, however at the end of the day, accuracy of the traffic information itself is what really matters.

I agree with Bryan that it is the accuracy of the data that matters most. The problem is that the same painting of colors on a traffic map can mean different things to different types of drives. Legends presented on various websites, traffic maps, and GPS devices doesn’t describe what it is we are looking at. If a road has a speed estimate of 25 during low traffic and cars are moving at 25 mph, what color should that be? If I’m a commuter and know the road it should be green. If I’m just traveling through the area I could probably deduce the speed of the road simply from the road classification. Should it still be green? If the legend says red=slow then should it be red?

I took a way too small of sample size, highly unscientific poll of a few people recently. I showed them a traffic map and asked them “Does the red line mean that traffic is moving slow (at a set speed) or does the red line mean that traffic is moving slower than expected?” The results were about 50/50. The next question to me of course was “which is the right answer?”. To which sadly I… me… the GPS guy… had to say “I honestly have no idea.”.

The Fix

So here is my proposition. Every traffic map should have two modes. the first mode is ‘Traveler’. In this mode traffic is color coded strictly on current flow data. Regardless of road classification, historical traffic data, or predicted traffic data– if vehicles are moving at set speed ranges, give them a particular color with a clearly defined legend.

A second mode should be called ‘Commuter’. Here, we really don’t care what the current speed/flow is. Simply display the traffic data according to how it differs from the predicted flow– either in comparison to road classification or (better) a historical traffic model. In this manner, green might mean that traffic is moving along at 90% or better of historical values for that road. If traffic is only moving at 50% of speeds for that day of week and that time color it red. And again– provide a clear legend so we know exactly what it is being presented.

TomTom Traffic

Incidentally, this “commuter” view is how the TomTom 740 LIVE was setup when I reviewed it– and thus the need for a detailed explanation on how they present traffic. In short, traffic was presented as “traffic” only if it differed from what their IQ Routes system predicted.

The Marketing Challenge

So why don’t we see maps setup from this “commuter” perspective? My guess is that it comes down to the impressions it gives a customer. When debates come up over which traffic provider is “better” the discussion almost always ends up looking at which map has more colored roads. The pressure to paint more colors on more roads is appealing to the traffic providers as it makes them look like they have more coverage in more places. To some extent that is true, but painting a traffic picture on a map and delivering good, actionable data are not necessarily in sync.

Traffic has been a hot topic lately with people debating which traffic providers (INRIX, NAVTEQ Traffic, Clear Channel, TrafficCast, AirSage, etc) provide the best data. Pretty darn impossible to tell when we can’t figure out exactly what it is being painted in various maps. In most cases, as with the Google Maps example, the colors simply represent current speed. I can make guesses of speeds simply by looking at the colors for road classifications. I don’t need a “traffic” map to tell me that.

I know why the sky is blue, but I don’t always know why this road is red.

13 Responses

  1. Tim, your quote “Incidentally, this “commuter” view is how the TomTom 740 LIVE was setup when I reviewed it– and thus the need for a detailed explanation on how they present traffic. In short, traffic was presented as “traffic” only if it differed from what their IQ Routes system predicted” is really only partially correct. While (covered) roads that are moving slower than IQR would predict, those same roads will show no display if moving FASTER than predicted. That would also be nice for a commuter to know. Now I have no idea how to implement that.

    gatorguy - September 21st, 2009
    • Good pont– though it isn’t as common for traffic to be moving faster than predicted and not many people will complain about arriving early. :) And what I said was still technically true… I mentioned when traffic was painted but didn’t talk about when it wasn’t painted. 😉

      Tim - September 21st, 2009
      • The reason it came to mind was that during the typical Friday afternoon rush hour, I275 in Tampa is normally jammed and I’ll go out of my way to avoid it. This past Friday? MSN was showing all green so I took a chance. It was clear sailing to my surprise. I agree it happens much less often, but in this case it was good to know since it ise the fastest way to one of my malls if there’s no traffic.

        gatorguy - September 21st, 2009
  2. Somehow I neglected to finish one sentence. While (covered) roads moving slower than IQR expects will be displayed on the traffic map, and roads moving within the range that IQR predicts will show no flow data, roads that are moving FASTER than expected will also show no flow data. And to be clear, I completely agree that much of what we see on traffic maps is not actionable. Too little information, or not clear exactly what is being displayed.

    gatorguy - September 21st, 2009
  3. Dash displayed live traffic in solid bands of color and historical (no live traffic) with a dotted line of color. There may be a solution in something like that.

    Scott G. Lewis - September 21st, 2009
    • That was decent solution, Scott. However it still doesn’t give you an indication how the live traffic compares with the historical. In other words if you saw a road indicating yellow current (not historical) traffic, is that flow normal for that time of day? That is what the commuter would want to know.

      Tim - September 21st, 2009
      • Scott didn’t say that was the exact solution, he just said “there may be a solution in something like that”. It’s not a huge leap to imagine having live traffic in solid color and historical traffic as a dotted line next to the live data. Admittedly this clutters up a small screen with what may be irrelevant information. As I just posted, I’m of the philosophy that I want my navigation system to be able to accurately select the fastest route. This is true if I’m a commuter or a visitor. I don’t care how it displays the data if the device does its job.

        Andrew - September 22nd, 2009
  4. On a navigation system, who cares how the current traffic compares to historical data? The system should use live traffic patterns to find the fastest route to the destination.

    Admittedly, in the past, there were some problems with both the quantity and quality of data. I was a Dash owner, and frequently ignored its recommendations in the early days if I was commuting.

    As data is increasingly collected directly from the users (connected GPS like the Garmin 1690, gps-enabled cell phones), both the quantity and quality of data will improve, and users will be able to trust their navigation systems to provide the fastest route.

    The only situation where this is a significant problem is when using google maps for looking at traffic patterns but not doing any routing. Google maps already–in some cases–can predict the transit time “in traffic”. I would imagine that as their data gets better, you’ll see the option in google maps to automatically choose the fastest route in live traffic.

    Andrew - September 22nd, 2009
    • Interesting thoughts Andrew, though I’m not sure I agree.

      On a navigation system, who cares how the current traffic compares to historical data?

      If that was the case, then why do GPS devices show traffic data at all? If the GPS will pick the best route through traffic, then why do the roads need to be colorized? I think the answer is that people don’t completely trust the data. Speaking from the eyes of a commuter, they want to see the data but then make their own decisions. If a commuter has a navigation system with a traffic map that colorized based on comparing current to historical traffic they might not even create a route to where they are doing– they might just look at the map, make their own routing decision, and drive. I’m not saying people should make their own decisions as the little processor can do the math easier, but people generally don’t trust the traffic data and tend to want to make their own decision… correctly or not.

      Tim - September 22nd, 2009
      • As I stated, there are existing problems with traffic data acquisition and distribution, creating a mistrust of the data. However, I think we are on the verge of that trust disappearing with connected GPS devices going mainstream.

        In the early days of online maps, there was a lack of trust in the route. In the early days of in-car GPS, there was the same problem. In both cases, the data has gotten better, and people are confident on relying on the provided route. We are in the early days of live traffic data. Google is now collecting live traffic data from cell phones. RIM has acquired Dash. Garmin’s 1690 is coming soon. Connected GPS devices will soon become ubiquitous, and people will start trusting the data.

        I understand your concerns. When I use google maps, it’s not very good at routing around bad traffic. When I had a Dash, I’d frequently use my own route as a commuter. The Dash didn’t have a data quality problem. Where it had data it was quite good. The problems were in data quantity and poor algorithms. In both cases, it was nice to be able to visualize the data and pick my own route. I just don’t think that will be necessary in the near future (within the next 1-2 years).

        Andrew - September 22nd, 2009
        • While I agree with you that we’re on the verge of just “trusting” our GPS, I don’t think it has anything to do with “connected” devices. A traffic antenna gives you just as much download capability as any connected device.

          I don’t think the upload capability of a GPS will help in any material way. GPSs aren’t ubiquitous enough to provide meaningful probe data to traffic vendors. The only usable scale is in probes from a cellco company like Airsage/Verizon.

          Once this cell-probe info is translated into traffic data, it can be broadcast via any medium: antenna, cell data, etc.

          mvl - September 26th, 2009
          • I should have been more clear in my follow-up post. When I refer to “connected GPS”, I am referring a device with GPS that has the capability to upload data (cell-enabled GPS or GPS-enabled cell phone). I listed both in my original post, and should have done the same in the follow-up. I agree that GPS-enabled cell phones will probably provide the bulk of the data, but to me the distinction is blurring and becoming irrelevant.

            FM and XM traffic information are limited by their simplex distribution methods to major metro areas. Due to the simplex nature of the channel, there may be a bandwidth problem. Let’s assume that flow data is better than incident data. Let’s also assume that the goal is to provide all of the relevant and reliable data to each traffic-enabled GPS device (be it a traditional navigation device or a cell phone). With a simplex distribution channel, the data source has no way of filtering the data to reduce the necessary bandwidth, beyond just limiting the data to major arteries or decreasing the depth of the data. It must broadcast all the data it has that is relevant to its coverage area, and it must broadcast on some sort of cycle, limiting the update frequency for the end-user. The user might end up being in the middle of slow traffic before the GPS hears about it from the broadcast source. There is a trade-off between the breadth of the data (just freeways, freeways+major surface streets, or all surface streets), the depth of the data (just incident data or incidents+flow data) and the update interval.

            With a duplex cell connection, the device has the ability to query the source, requesting only the relevant data and receiving the relevant data faster. Since irrelevant data is filtered out, there is room on the channel to include better information: more surface street data and traffic flow data wherever it’s available.

            If you only need data for major arteries AND only need data for metro areas AND can tolerate delays of 15 minutes AND can tolerate only having incident data in some areas, then there is no distinction between a traffic antenna and cell data traffic.

            Personally, I’m looking forward to the 1690 and similar devices.

            Andrew - September 28th, 2009
  5. Tim those Google traffic color codes are outdated and inaccurate. Note that the green color is often displayed on small secondary roads where the speed limit and actual speeds are under 30mph. That alone invalidates Google’s last update on that. The TT codes are horrible and difficult to fathom. I prefer seeing black, red, orange and green on the road ahead of me as I don’t always have a set destination running. The Garmin literature says the 1690 will provide that.

    As traffic probes move to using the gps in consumer cell phones traffic reporting accuracy will increase exponentially.

    Inrix has long been the leader in fleet probes citing about 900,000 of those probes. Of those some percentage travels at night to avoid traffic, some percentage is rarely on secondary roads, some percentage is dropping off a load for quite some time and not feeding back and some percentage of fleet vehicle probes are good only for historical data since the data is delayed. At the end of the day they never had 900,000 probes available and at any one time much less than 900,000.

    Mistelle is po’d because the competition, here Google, is giving away what he wants to charge for making Inrix less valuable and maybe in trouble. At the other side of him is Navteq Traffic with the huge resources of Nokiq about to eat Inrix for Lunch here and in Europe. The Darkhorse here is NIM which has 3 million gps probes and moved to use them by buying the back office ops of TrafficGauge. That could tip the probe count depending on who gets access to them.

    The 740 is good at avoiding traffic jams I anticipate that the 1690 will be better but time will tell and these are really still first generation real time devices. Dash had a good concept but rather than focus on the core routing and traffic avoidance they made it the darling toy of the blogosphere and ended a commercial failure.

    Jim - September 24th, 2009

content bottomleft content bottomright