THE CONTENT YOU WANT. NOW.
Enjoy our lighthearted views on the camera dataset market. If you don't enjoy sarcasm - skip it:
No, all traffic camera datasets are not created equal.
There are two types of datasets available: those that are built by user contributions and those that are not. User-generated content sites (UGC) are a dime a dozen and can be found all over the Internet.
UGC sites do some things well: they put up a great website. Some of them make some nice software. Others companies promote their UGC sites to try to sell imported electronic devices. What they don't do well is develop datasets. In fact, many of the UGC sites can't even tell you what is in their data. They let the user community develop it for them. This helps the dataset develop faster they claim. Hardly.
Some Facts about UGC Datasets:
Think about it. There are at least 5 major UGC datasets available on the Internet. What percentage of the driving public do you think participates in these websites? Yeah, it's low. And further, what percentage of those that decide to participate take the time to contribute to each of the 5 datasets equally? Yeah, it's really low. What do you end up with? You end up with a dataset accuracy that depends on the popularity of the website! Is that what you want when you are trying to avoid a costly ticket? Nope.
Is more data better? The myth behind fancy data attributes.
Some of the datasets available are the electronic equivalent of the swiss army knife. They come with all kinds of fancy attributes: direction of travel, speed trigger, etc. This is great, right? Nope. These fancy attributes come from one place: UGC. That's great for an initial installation, but how do they update the attributes when they change? They don't. Our data comes from a multitude of verified sources (nope we aren't telling!). Few of those sources provide these extended attributes. And fewer yet, publish them when they change. Cities all of the Country add cameras to an existing intersection, move them to a different approach when the revenue doesn't meet expectations the first time, add speed detection once the local ordinance or State law allows, change the trigger speed, etc. Recently, one municipality added "illegal left-turn" detection to their existing cameras. Datasets that attempt to classify all of these extended attributes can't keep up. And that means poor accuracy.
Great Idea! Yet another distraction in the car when you should be driving.
Otherwise known as "Don't filter your alerts!" The supposed selling point of these fancy extended attributes is then you can only receive the alerts that you want. Great theory. When your GPS alerts you, what is your natural reaction? You tap the brakes, right? That is exactly what we want you to do. Tap the brakes and be alert (from our alert!). Pretend you are using a fancy (but inaccurate) extended attributes dataset and after the alert you are shown text that tells you one of the following: red light camera, speed camera, red light and speed camera, illegal left-turn camera or oversized truck camera (yes they are coming). What did you do when you heard the initial alert? You tapped your brakes, right? That is the same thing you'll do with our dataset. You do that because it is a REACTION. It is instinct. You don't have time to read an electronic display and then mentally process "hmmm, what should I do" and then react. You already reacted.
Here is the problem. First - these are dangerous intersections - that is why the cameras are there to begin with. Do we want you focusing on reading a tiny electronic gadget several seconds before you enter these intersections? No. We want you focused on the road. What are you going to do differently knowing the type of alert it is? Nothing. Your instincts have already reacted. You tapped the brakes.
Second problem - what happens when the extended attributes aren't right (no - does that really happen?). Last month the intersection was a red light camera. The State approved a law that allows speed cameras, the manufacturer of the camera has an available hardware upgrade and that intersection now detects red light and speed violations. Your new electronic gadget with the fancy extended attribute dataset doesn't alert you to the fact that you are going over the speed limit because it doesn't think it is a speed camera. Why? Because a user that feels like updating the UGC dataset hasn't been ticketed there for speed yet. (Don't you just love having to rely on users for accuracy?) But guess what - since you just blew through the green light at +10 mph, you just got ticketed. So now you can do your good deed to society and update 5 different UGC datasets online. Thank you for your contribution.
Folks - a good dataset should do one thing very, very well. It should alert you to an upcoming traffic camera. You will be alerted by our dataset and when you do, you should consider it a "Traffic Camera Alert." What does that mean? It means follow traffic laws. Be alert, slow down, don't run a red light, don't make an illegal turn. In the next several seconds, DO EVERYTHING RIGHT. No reading a screen. No deconstructing the code of blinking LED lights. No listening for 14 different tones. Your instincts already hit the brakes for you - now just focus on the road and do everything right.
The importance of verification - you gotta love 'em.
One of our competitors, bless their hearts, has a website that preaches the importance of verification. We couldn't agree more. (According to them, anyone else building a dataset is a hobbyist and apparently hobbyists aren't able to verify data.) But wouldn't you know it, right after preaching about the importance of data verification, they are marketing their dataset with - you guessed it - fancy extended attributes like "dangerous intersections."
Here's one for you....How does one go about independently verifying a "dangerous intersection?" Is there some official government repository of dangerous intersections? Did John Smith, Marketing Guru, see a crash at Main St & Insurance Road and say "add that to the database!" Does it take 10 crashes, 100 crashes, 1000 crashes before an intersection is "dangerous?" What about 10,000 "near misses" but only 10 "hits." Is that dangerous? Here is the bottom line folks - don't just read marketing - read through marketing. Does a claim that dangerous intersections are independently verified pass the logic test? Nope.
But we've saved the best for last. Speed Traps. Yes, we've learned through reading marketing that speed traps can be independently verified. Cool! But, come on, really folks? You believed that for a minute, didn't you? How does one go about verifying speed traps? I was cruising down the Interstate once and I saw a police officer parked in the median. Is that a speed trap? Perhaps. I haven't seen him there since (it was 2 years ago). Is it still a speed trap? Can anyone come up with a credible model for the independent verification of a speed trap? Or perhaps is it more likely that we have stumbled upon a behind-the-scenes partnership with a (you guessed it) user-generated-content data provider? Frankly I don't know - but what does logic tell you? Don't believe the hype. THINK.
Where does their data come from anyway?
Take a good hard look at the data contained in some of these UGC websites floating around. Our dataset is the only dataset that is built with multiple definitions at complex intersections. This gives you the maximum alert warning without losing some of your alert radius to the size of the intersection itself. The added complexity also makes it easy to spot our data in the wild. It would take a huge investment of time for someone to purge the data of these multiple definitions. And we all know the UGC sites don't like to spend much time on the data. We count at least 3 people behind various imitation datasets as our customers! We have some pretty good detectives! We just can't figure out why they used @hotmail.com or @yahoo.com email addresses and PO Box street addresses? Seriously though folks, if you find data with multiple definitions at the same intersection, you can be pretty confident that you aren't buying the data from the original source. Come on over and buy the data direct from the source and ensure you are getting the most timely and accurate data available today.
No, all traffic camera datasets are not created equal.
There are two types of datasets available: those that are built by user contributions and those that are not. User-generated content sites (UGC) are a dime a dozen and can be found all over the Internet.
UGC sites do some things well: they put up a great website. Some of them make some nice software. Others companies promote their UGC sites to try to sell imported electronic devices. What they don't do well is develop datasets. In fact, many of the UGC sites can't even tell you what is in their data. They let the user community develop it for them. This helps the dataset develop faster they claim. Hardly.
Some Facts about UGC Datasets:
- Inaccuracies abound! Because we have been selling our dataset to the public since 2006, we know that the typical consumer doesn't necessarily have a deep knowledge of traffic cameras, nor should they. We routinely receive email from customers claiming they saw a camera when in fact it is a stoplight traffic flow camera, or a DOT traffic observation camera. We graciously thank the customer and try to explain the various cameras types that exist, but the incorrect location doesn't make it into the dataset. In a UGC site, that location gets entered by the consumer. No questions asked!
- Removal of cameras
. Check out some of these UGC websites yourself. Do you see a way to remove a camera that has been relocated or removed? We don't either. You can rate it "thumbs down." What does thumbs down mean? The camera is either there or it isn't.- Notoriously dated
. After all, it takes the camera to be installed and a user of the website to drive by it. Then that user has to take the time to enter it in the database. How many people have already been ticketed before someone decides to do that?
Think about it. There are at least 5 major UGC datasets available on the Internet. What percentage of the driving public do you think participates in these websites? Yeah, it's low. And further, what percentage of those that decide to participate take the time to contribute to each of the 5 datasets equally? Yeah, it's really low. What do you end up with? You end up with a dataset accuracy that depends on the popularity of the website! Is that what you want when you are trying to avoid a costly ticket? Nope.
Is more data better? The myth behind fancy data attributes.
Some of the datasets available are the electronic equivalent of the swiss army knife. They come with all kinds of fancy attributes: direction of travel, speed trigger, etc. This is great, right? Nope. These fancy attributes come from one place: UGC. That's great for an initial installation, but how do they update the attributes when they change? They don't. Our data comes from a multitude of verified sources (nope we aren't telling!). Few of those sources provide these extended attributes. And fewer yet, publish them when they change. Cities all of the Country add cameras to an existing intersection, move them to a different approach when the revenue doesn't meet expectations the first time, add speed detection once the local ordinance or State law allows, change the trigger speed, etc. Recently, one municipality added "illegal left-turn" detection to their existing cameras. Datasets that attempt to classify all of these extended attributes can't keep up. And that means poor accuracy.
Great Idea! Yet another distraction in the car when you should be driving.
Otherwise known as "Don't filter your alerts!" The supposed selling point of these fancy extended attributes is then you can only receive the alerts that you want. Great theory. When your GPS alerts you, what is your natural reaction? You tap the brakes, right? That is exactly what we want you to do. Tap the brakes and be alert (from our alert!). Pretend you are using a fancy (but inaccurate) extended attributes dataset and after the alert you are shown text that tells you one of the following: red light camera, speed camera, red light and speed camera, illegal left-turn camera or oversized truck camera (yes they are coming). What did you do when you heard the initial alert? You tapped your brakes, right? That is the same thing you'll do with our dataset. You do that because it is a REACTION. It is instinct. You don't have time to read an electronic display and then mentally process "hmmm, what should I do" and then react. You already reacted.
Here is the problem. First - these are dangerous intersections - that is why the cameras are there to begin with. Do we want you focusing on reading a tiny electronic gadget several seconds before you enter these intersections? No. We want you focused on the road. What are you going to do differently knowing the type of alert it is? Nothing. Your instincts have already reacted. You tapped the brakes.
Second problem - what happens when the extended attributes aren't right (no - does that really happen?). Last month the intersection was a red light camera. The State approved a law that allows speed cameras, the manufacturer of the camera has an available hardware upgrade and that intersection now detects red light and speed violations. Your new electronic gadget with the fancy extended attribute dataset doesn't alert you to the fact that you are going over the speed limit because it doesn't think it is a speed camera. Why? Because a user that feels like updating the UGC dataset hasn't been ticketed there for speed yet. (Don't you just love having to rely on users for accuracy?) But guess what - since you just blew through the green light at +10 mph, you just got ticketed. So now you can do your good deed to society and update 5 different UGC datasets online. Thank you for your contribution.
Folks - a good dataset should do one thing very, very well. It should alert you to an upcoming traffic camera. You will be alerted by our dataset and when you do, you should consider it a "Traffic Camera Alert." What does that mean? It means follow traffic laws. Be alert, slow down, don't run a red light, don't make an illegal turn. In the next several seconds, DO EVERYTHING RIGHT. No reading a screen. No deconstructing the code of blinking LED lights. No listening for 14 different tones. Your instincts already hit the brakes for you - now just focus on the road and do everything right.
The importance of verification - you gotta love 'em.
One of our competitors, bless their hearts, has a website that preaches the importance of verification. We couldn't agree more. (According to them, anyone else building a dataset is a hobbyist and apparently hobbyists aren't able to verify data.) But wouldn't you know it, right after preaching about the importance of data verification, they are marketing their dataset with - you guessed it - fancy extended attributes like "dangerous intersections."
Here's one for you....How does one go about independently verifying a "dangerous intersection?" Is there some official government repository of dangerous intersections? Did John Smith, Marketing Guru, see a crash at Main St & Insurance Road and say "add that to the database!" Does it take 10 crashes, 100 crashes, 1000 crashes before an intersection is "dangerous?" What about 10,000 "near misses" but only 10 "hits." Is that dangerous? Here is the bottom line folks - don't just read marketing - read through marketing. Does a claim that dangerous intersections are independently verified pass the logic test? Nope.
But we've saved the best for last. Speed Traps. Yes, we've learned through reading marketing that speed traps can be independently verified. Cool! But, come on, really folks? You believed that for a minute, didn't you? How does one go about verifying speed traps? I was cruising down the Interstate once and I saw a police officer parked in the median. Is that a speed trap? Perhaps. I haven't seen him there since (it was 2 years ago). Is it still a speed trap? Can anyone come up with a credible model for the independent verification of a speed trap? Or perhaps is it more likely that we have stumbled upon a behind-the-scenes partnership with a (you guessed it) user-generated-content data provider? Frankly I don't know - but what does logic tell you? Don't believe the hype. THINK.
Where does their data come from anyway?
Take a good hard look at the data contained in some of these UGC websites floating around. Our dataset is the only dataset that is built with multiple definitions at complex intersections. This gives you the maximum alert warning without losing some of your alert radius to the size of the intersection itself. The added complexity also makes it easy to spot our data in the wild. It would take a huge investment of time for someone to purge the data of these multiple definitions. And we all know the UGC sites don't like to spend much time on the data. We count at least 3 people behind various imitation datasets as our customers! We have some pretty good detectives! We just can't figure out why they used @hotmail.com or @yahoo.com email addresses and PO Box street addresses? Seriously though folks, if you find data with multiple definitions at the same intersection, you can be pretty confident that you aren't buying the data from the original source. Come on over and buy the data direct from the source and ensure you are getting the most timely and accurate data available today.
2010 Copyright GPS POI US, LLC
