Saint Petersburg, FL
PC Mag had it right...Oookla had it wrong...
I'll catch all sorts of hell for this but I know I'm right and it needs to be said. PC Mag had the general idea right to begin with, Ookla was wrong and they have always been wrong.
First of all the online user experience should NOT be measured by web browsing alone because being online is far more than just surfing nowadays.
There are two major types of measurements that should be made when you are benchmarking the quality of your internet connection.
1) Application Speed and Quality - The absolute maximum speed and quality of service obtainable through a single socket test. This is important because streaming, voip and similar applications in most cases are NOT multi-socket but are in fact single socket. That said this form of testing is critical in evaluating the real quality of an internet connection because for example if your subscribed tier is 40/5 and your single socket application test results rates you at 5/1 with a QoS of 37% you are going to have serious problems with certain applications.
This happens a lot with certain providers and yet very few benchmarking sites help users diagnose these issues when this is what people should be looking at. For most windows based systems the above referenced connection will top out somewhere around 12-15/4.8 with a QoS of not less than 90% when running a single socket TCP test to be considered a good connection. There are ways to improve this depending on your operating system, distance between test site and obviously the quality of your network connection.
2) Capacity - This is more like the traditional speed tests which first calibrate based on connection speed, determining at what datarate the connections starts to drop packets and then open a specific number of sockets based on the calibration speed for the duration of the test. This test should ideally also measure QoS (or the connections ability to maintain a steady datarate for the duration of the test). This is the test that essentially behaves like modern browsers would.
It's critically important to know the results of both tests before making an evaluation of the connection quality and Ookla doesn't offer you both of these tests so any results they provided PC Mag would have been only marginally useful.
PC Mag was on the right track to begin with...they just did an awful job of explaining it to the public and press who apparently hassled them over it and apparently caved this year to the pressure...sad really because they had an opportunity to properly educate the public on internet connection benchmarking and blew it.