Broadband.gov
Federal Communications Commission



Consumer Broadband Test Update

March 17th, 2010 by Jordan Usdan - Acting Director, Public-Private Initiatives

Thanks to the over 150,000 unique users who have taken over 300,000 Consumer Broadband Tests, as well as the nearly 4,000 addresses submitted to the broadband Dead Zone Report. The popularity of the consumer tools has exceeded our expectations.We’ve made some text changes to the short “About” section found on a tab below the Consumer Broadband Test Tool. Some users have been confused by the differences between the two testing platforms presented by the FCC – Ookla and M-Lab – and this section explains the variability

Over the weekend, the FCC also updated both the Android and iPhone FCC Apps to improve the user experience.  The FCC App can be found by searching for “FCC” in either the Android or iPhone App store.

The FCC chose to use two testing applications for the Beta version of the Consumer Broadband Test.  The two applications are among the most popular on the Internet and the FCC hopes to make available additional testing platforms in the future. However, software based broadband testing is not an exact science and contains inherent variability, as described in the About section.  This is why the FCC will also be conducting a hardware based scientific study of broadband quality across the country.  See this recent blog post about this venture, and the RFQ here.  The FCC will use the results of this hardware study for analytical purposes. The results of the software bases testing (see data below) are interesting and show broad trends, but the FCC is not relying on the data for analytical purposes.

Here are the user experienced differences between the two testing platforms:

 

 
Metric
M-LAB
OOKLA
 Average Download Speed (mbps)
7.04
11.5
 Median Download Speed (mbps)
3.95
8.14
 Average Upload Speed (mbps)
2.74
2.09
 Median Upload Speed (mbps)
0.87
1.01
 
You will see that Ookla provides a higher overall average and median speed than M-Lab.  This is likely due to the different methodologies these testing applications use.  The difference comes from the fact that broadband speeds vary over time, even within a single second. Ookla measures peak performance and ignores short periods of slow speed, which it considers to be speed bumps in performance, while M-Lab takes many rapid speed measurements and averages them all. For more detail, see the Ookla and M-Lab methodology sections.  Additionally, Ookla and M-Lab each have testing servers geographically distributed across the country.  Individual’s proximity to these testing servers could also affect testing results.
 
Although software based testing cannot provide users with a 100% reliable measures of broadband quality, the FCC makes these tools available as they provide comparative and relative real-time performance information and helps the FCC collect broadband availability data.
 
Here are some interesting data and maps from the first six days of the Consumer Broadband Test. This data is derived from the results of both testing applications.
 
 
As you can see, 87% of test takers are home users, which is the FCC’s target audience with this application. Additionally, a clear trend is visible across business sizes, high bandwidth connectivity for community institutions, and lower bandwidth for mobile connections. Again, these results are non-scientific extrapolations from the Beta version of the Consumer Broadband test. Additionally, about 98% of user submitted addresses are geo-coding correctly, which is a very good rate.
 
Given that this is the Beta version, we want to hear from you about additional features we can add to this interface.  We already have some internal plans to rollout an updated version in the near future that provides greater context to users about the meaning of their testing results.  So please reply to this blog with your suggestions!

30 Responses to “Consumer Broadband Test Update”

  1. Guest says:

    just download something from usenet and you will get your speed. I have 3 megabit and i download at 320-340 KB/s all the time.

  2. Guest says:

    Man you guys suck and your speed test is not accurate. I don't know why you deleted my last comment.
    usenet is the only speed test I'll ever need. Ive had many isp providers and Iv always got what they said I would except for dsl but thats to be excpected from dsl. If you have cable you should get what your supposed too and if you think your not its most likely because your running tests like these.

  3. David Hoffman says:

    On the Consumer Broadband Test Origins slide could you put the percentages and numbers by the categories. The small slivers on the pie chart are difficult to distinguish. You may want to go to a solid color for each slice. The graduated colors are difficult to distinguish. Same thing with the bar charts. Thanks for the great explanation of the two tests. I was getting a high reading for download from one that did not make sense compared to speedtest.net. I now understand why. Thanks for a great start to getting a true picture of what data rates users are really getting.

  4. Guest says:

    No data on any of the US Territories? Guam? Puerto Rico? USVI?

  5. Dave Burstein says:

    It's so rare to see thoughtful analysis of data like this it's a pleasure. I may use the "fixed over the weekend" in my reporting. That's a new spirit at the FCC I saw at the broadband plan.

  6. Guest says:

    @8:35

    Uh, the dot com bubble happened near the end of the Clinton administration. The bubble was created in the late 90s by speculating over the sucess of companies run by college kids who could design a fancy website but didn't know the first thing about running a business. Companies that were run well like Google, Amazon, and eBay survived.

  7. Philip Kahn says:

    You'll probably have more readable graphs (well, aside from that color thing) if you save them as PNG files, rather than JPG files. Possibly (though not definitively) smaller, too.

  8. Guest says:

    This is the first I heard of a national broadband test. Some representative sampling. I would have liked to have tested my diinosaur dial up speeds. Still can't get broadband where I live due to ComCast's inability to live up to their contract with my town. Yet 3 telephone poles away I can see the terminus of TimeWarner's cable.

    We need a national plan, like the telephone act of the 1930's (or was it called the Rural electrification act) Leaving it to corporations and their hunger for the "bottom line" has put our country decades behind the rest of the world.

    For that I do blame the previous administration as the one prior to that brought us the dot com bubble. All that innovation did create new technologies, before it burst. Had an admin come in that knew how to "use the google" and didn't lose all it's e-mails we might have had FIOS to every dairy barn by now.

  9. Stephen says:

    Some suggestions

    * Larger files (and/or multiple files of different sizes) to be downloaded to avoid ISP speed bursts skewing the results.
    * Different types of files; not just compressed files, but things that people actually download.
    * Several different servers, preferably revolving, to help avoid ISPs' deliberately pumping connections that are made to your servers. It happens with every benchmark (sometimes via drivers and sometimes through "optimizations") and it will probably happen here. When I think of the money that's involved here, well, let's be realistic.

    * IPv6 tests, if possible, to see who can actually use it because we're all going to need to sooner rather than later.

    In regards to the maps, please change their colors if you can. The colors all look too similar to my eyes (but I have a slight color deficiency). Also, the maps look very rough. That makes them a little more difficult to read than they need to be, but I can understand compressing their file sizes.

    I like what you're doing so far. You have a "go get 'em" attitude, and I love it. Frankly it's about time. Just stick to your guns and make this happen.

    I talk to a guy every day who's stuck on dial-up. AT&T began installing new lines for DSL in his area and then they just stopped. No explanation for it, they just stopped and he can't get anywhere with them, and of course there's no competition in his area (does anyone have serious broadband competition available to them?). That was over a year ago, as I recall.

    Do not leave this to corporations. Don't take deals or promises any longer. They have had more than a fair chance. You need to actually do something, so I hope that you do. We'll see.

  10. J Melendy says:

    I think that providing users of the test with localized results would grow increasingly interesting as the database of results grows. It is already revealing to see the results broken down by state. As a rural town user of broadband, the local maximum advertised speeds are less than the reported average for the state. This makes me very curious to see the correlation between the providers and available speeds in rural areas in my region.

  11. Guest says:

    I get 25 mbps upload from speedtest.net, but only 5mbps from this test. I did the tests back to back more than once. Someone may want to look into that....

  12. Joe K says:

    With Ookla consistently reporting download speeds around three times those reported by MLAB and most other speed checkers available on the Internet, and far above the maximum data rate offered by the ISPs to their customers, it seems to me your speed graphs, if not corrected for this anomoly, are uselessly inaccurate. My own comparisons of these tests were run from Iowa, Arizona and Florida and the results are consistent.

  13. Guest says:

    Take the Test here:

    http://broadband.gov/

  14. James Williams says:

    I have Time Warner as my internet provider, and they advertise this thing called power boost. Basically when I start downloading something, it gives me a large amount of bandwidth (up to around 15 Mb/s) for a few seconds, then it drops down to around the usual 7 Mb/s. The problem (and I would guess the real reason for having this technology) is that it skews speed tests like this one in their favor. They may only provide 7 Mb of bandwidth, but when I run various speed tests I routinely see numbers above 10 Mb/s. It's deceptive on their part, and I wonder if these types of gimmicks are being taken into account. When I ran the FCC's broadband speed test, it said my download speed was 12Mb/s, and I know that I don't get that kind of speed through normal use from Time Warner.

  15. Sam says:

    I think PNG or 8bit GIF format will reduce the size of images a lot.
    Any way the speed test is very helpful service and more reliable than others.

    Sam
    <a href="http://www.minutesofmeeting.net">pool paint</a>

  16. Mace Moneta says:

    It would also be interesting to see the bandwidth breakdown by local time of day.

    In my area, bandwidth frequently drops to as low as 25% of normal during the day, and goes to full bandwidth after midnight. That's with multiple concurrent destination site testing, so no "the server must be slow" excuse.

    There's also an issue with uptime. 98% uptime seems good here, with some outages lasting many hours. That's not a problem for most web sites, but it's a serious issue for those that use VoIP services for emergency calls.

  17. Guest says:

    If you aren't already looking at it, you might check out Myspeed Connection Server from Visualware as they support true TCP connection testing unlike Ookla which seems to just do a "data blast" to saturate the connection. Java also seems to perform better for tests over 50 Mbps. I've had trouble getting higher speeds on multiple flash based tests.

  18. Guest says:

    1=53000 bits per second, 2=512000 bps, 3=1000000 bps, 4=10000000 bps, 5=100000000 bps

    The fact we're even reading this indicates we understand the above. We're debating the fact that #4 or #5 may not be sufficient for current needs. Understandable. We're debating the technology versus market penetration versus cost to implement #4,#5 and #6, #7 whatever they may be. Understandable. Let us all agree this debate is UNDERSTANDABLE.
    To the people stuck at #1, #2 or #3 is broadband. Provide that. In 6 or 12 months they'll be here to join the debate. To the people that don't even own a computer, provide a simple means of connecting it to the world and they'll be here in 12 or 18 months also. The debate will rage on among people that understand a little more about something they didn't understand 18 months ago. What idea's could possibly be lurking in the minds of isolated individuals out there? Hopefully a way to provide cost effective multi-exabyte backhauls to keep up with the increased demand for MORE!

    -Ohio

  19. Guest says:

    these tests shoshed my down speed ~megabits/s when in real life it is kilobits/s under the best scenario... Oh and do not remind me about week-end speeds. So either Optimum online does prioritize the traffic for these tests either, somebody else does, or this just coincidence... you pick. Anyway these tests do not reflect reality of my little universe

  20. Warren.Lambert says:

    In my experience consumers generally don't distinguish mega-Bytes per second and mega-bits per second unless they are technicians. Most wouldn't distinguish b and B, as in mbps and mBps. Adding kilo bits per second and kilo Bytes per second adds to the confusion.

    I hope broadband.gov provides a consistency of units, always using the same measure, to clarify this minor point. Perhaps you could handle this as considerate writers handle English-metric, e.g. '39.4 inches (1 meter)'. Footnotes also might help to distinguish kilo-mega and bits-Bytes.

    The public won't understand data rates until they have a clear idea about whether their WWW speed is fast or slow. If the units are mixed, most will remain totally confused.

  21. Carlo says:

    What a great Idea with this broadband initiative , I would also like to suggest to use solid colors instead of the current scheme, overall its a great start and looking forward to you next tests , keep up the great work and remember Rome wasn't built in a day , important and well-thought out ideas take time and patience , unfortunately that is something that is lacking in our current hurried society. Its about time our country jumps aboard the 21st century track , keep up the great work !!!!!

  22. Guest says:

    Okla sites are faster or slower based upon the proximity to those being tested and the load on that test server at that time. None of the tests we performed were representative of the actual speeds we normally see. If we go to four random Okla (speedtest.net) sites, we get four different sets of numbers. Since they are using Okla, the broadband.gov speed test does not provide quality, believable data.

    Too bad they didn't use the Web100 based Network Diagnostic Tool (NDT) speedtest servers. Maybe they should call them and ask for some assistance in collecting meaningful data.

  23. Ryan Jairam says:

    This needs to be broken down a bit more. By state is not enough, because in some states there is a huge imbalance. For example, in my home state of New Jersey, some areas have access to Verizon FiOS and Cablevision Optimum Online which gives download speeds up to 50 and 100Mbps and upload speeds up to 35 and 15Mbps respectively. Meanwhile many others don't have access to either of those providers and our upload speed is stuck at 2Mbps or less. I live in Sussex County, Verizon doesn't serve here and the fastest we can get right now for any residential connection is 30Mbps down, 2Mbps up from Service Electric Cable/PenTeleData. That tier costs a whopping $70. Most people are on 10Mbps or less download (with only 1Mbps up), with a huge number on 5Mbps from either the cable (Service Electric) or the phone company (CenturyLink). In short, I am happy with the download speed but the upload speed is seriously lacking, but there is little the companies desire to do about it because the perception is that high upload speeds aren't really needed for residential connections. I do some of my work from home sometimes and I often need to upload large files. I'm sure many others would like higher upload speeds as well. We're also constrained by a bandwidth cap that is as low as 20Mbps for some users and 100Mbps for the highest tier. Other areas have 250GB caps and higher. This is one thing that needs to be addressed as well.

  24. bob greer says:

    please use the results to get the isp to give us the downloads speeds they charge for, my starband.net service has never given the top speed they sell!

    even i burst mode !!
    they should have to realiy give at least within 15 % of the speed 90 percent of the time !!
    thanks
    for this site !!
    bob

  25. Drewthis says:

    I have service by Time Warner at home. The first time I checked using your tests these are the results I received.

    Mlab

    Download speed - 1057
    upload speed - - - - 367
    latency- - - - - - - - - 135
    jitter - - - - - - - - - - 117

    OOKLA

    Download speed - 4965
    upload speed - - - - 371
    latency- - - - - - - - - 40
    jitter - - - - - - - - - - - 1


    Several days later when I tested again I could only reach the OOKLA tests. I run a number of websites, and I know a little bit about redirecting. I think you guys were gamed by Time Warner. You might want to look into that.

  26. Brian says:

    I just took this test and it came really close to the max limit speeds that I am paying for from my ISP.

    I also did a back to back test with speedtest.net and it only came with 1/2 of what this test claimed I got.

    I would think that since both of those tests are powered by ookla that they would be fairly close. I am not a networking guy but to me that sounds like my ISP has picked up on this test and is doing some prioritizing of their packets.

  27. hondo says:

    None of these test are testing the speed of the connection that you are purchasing from your Internet provider. All of these test and dozens of other tests, especially the SpeedMatters.org test, are showing the end-to-end through put speed. If a customer is not getting the speed they are paying for in their connection to their provider the slow-down is going to be 99% of the time due to inefficencies in their computer, the multi-handoffs from one world-wide-web provider to another in the connecting server, network congestion 100s of miles from the customers location, congestion at the server at the far end, etc.

  28. Evan says:

    My biggest request would be for some honesty in pricing regulations - especially in the voice area but also in broadband - you should be able to get exactly what you are paying for, and hear the real price when the salesman calls, not find $30 per month inflated to $50 by nickle and dime extra charges when you get the bill.

  29. Atishay jain says:

    This is the first I heard of a national broadband test. Some representative sampling. I would have liked to have tested my diinosaur dial up speeds. Still can't get broadband where I live due to ComCast's inability to live up to their contract with my town. Yet 3 telephone poles away I can see the terminus of TimeWarner's cable.

  30. Jet Ski Performance Parts says:

    Being consistent is very much important to this broadband test.

Leave a Reply



Capture The Phone Numbers Using Your Camera Phone

If you have a camera and a 2D matrix code reader on your mobile phone, you can capture the FCC Phone numbers right to your phone by following these three easy steps:
Step 1: Take a photograph of one of the codes below using the camera on your mobile phone.
Step 2: Use your phone's Datamatrix or QR Code reader to decode the information on the photograph. Please note, these code readers are device specific and are available to download on the internet.
Step 3: Store the decoded address information to your phone's address book and use it with your Maps or GPS application.

Datamatrix and QR FCC Phones