Hey Everyone, Good News! Verizon, AT&T, T-Mobile, and Sprint are able to deliver increasingly faster downstream speeds to your smartphone. The question is, how reliable are those speeds? Can you bank on getting 50 megabits per second every time you try to download a large photo to your smartphone?
The graph below shows the maximum downstream speeds measured for each of the four providers over the life of our mobile field testing program. During this multi-year period, we've seen the introduction of LTE networks by all four providers, and just recently Verizon announced the introduction of LTE Advanced, which promises a 50% increase in speeds.
Newer technology = faster speeds.
Those of you familiar with this blog know that faster speeds sure are great, but more important is the reliability of those faster speeds. How likely will I be able to experience the faster speeds consistently?
One metric we collect is the standard deviation. A low standard deviation suggests that the speed measurements tend to be near the mean (average). Conversely, a high standard deviation suggests that the speed measurements are spread out over a wider range of values from the mean. Higher standard deviation suggests wider fluctuations in speeds.
The data we've collected suggest that as download speeds get faster, the standard deviation increases. For one session, you might get 2 megabits per second, and another session, 24.
Here is a graph comparing the mean downstream speeds between Fall 2015 and Spring 2016. AT&T, T-Mobile, and Verizon seem to be delivering faster average downstream speeds compared to six months before.
Now, let's look at the standard deviation as a percentage of the downstream speed. Standard deviations have increased as a percentage of the average downstream speeds.
What does this mean? It's hard to say, but it appears the likelihood of getting the average speed at a particular location is lower than before. Here is what the average downstream speeds look like after lowering the mean downstream speeds by two standard deviations. Using this approach, we see a drop rather than an increase compared to the previous round of testing.