LTE take-up in the United States has been impressive since Metro PCS rolled out the first LTE network in September 2010, with around 50% of US OpenSignal users on LTE contracts. The four main US Networks are not uniform in their performance, and in this report we focus on the differences between them with regard to both speed and coverage.
The challenge when it comes to evaluating data networks, and LTE networks in particular, is how to weigh coverage and data performance when looking at a network as a whole. LTE has delivered vastly improved speeds on existing 3G and HSPA+ technologies, even though the speeds experienced by US users are much slower than most other advanced LTE networks worldwide. What is arguably more important, however, is the extent to which users are actually able to access the network – which we measure by recording the average user’s time with access to LTE, or ‘Time on LTE'.
It is important to remember that LTE performance isn’t a summary of a network’s overall performance – as 3G still remains extremely important, both to users with LTE and without. Around 50% of OpenSignal users in the US have LTE access and those users spend an average of 67% on LTE – meaning that they are still on 3G a good proportion of the time.
T-Mobile are comfortably the fastest network in the US over the last 3 months, with average speeds of 11.5 Mbps. Sprint perform worst of all US networks, posting LTE speeds that are scarcely faster than existing HSPA+ technologies, with their average speed of 4.3Mbps ranking as one of the slowest networks worldwide.
For comparison we measured Australia as the fastest country worldwide at the end of 2013, with speeds averaging 24.5 Mbps – which shows how much slower average LTE speeds are in the US when compared to their global equivalents, with the USA ranked 15th of the 16 countries we surveyed worldwide.
This metric is our unique way of looking at coverage, based on user experience rather than geography. Coverage is most important where users actually spend their time, especially for LTE (as it provides a layer of service that is non-essential for emergencies, unlike voice or basic internet connection). The ‘Time on LTE’ metric looks at the proportion of time a user has access to the LTE network, which gives a more accurate look at how real-world users are being served by their connection.
For the US we see Verizon performing best for this metric, with the average user having access to LTE 83.2% of the time over the last 3 months. Sprint again perform worst for this metric, with users having LTE access only 56.5% of the time, with T-Mobile only performing slightly better – meaning that their users experience high speeds but comparatively infrequently. In terms of an international comparison, the USA ranked 6th out of the 16 surveyed nations, considerably better than they perform for speeds.
Mobile networks are not constant; their performance fluctuates over time. Operators are constantly upgrading their technologies and rolling out to new areas, while increased user load on LTE networks combats these improvements.
Consequently it can be seen that all four networks trend downwards over the past year, with a reasonable amount of fluctuation from month-to-month. T-Mobile’s rise in performance in March/April last year marks the beginning of their public rollout (although they had been testing it comprehensively in advance of that) and the fluctuation is a result of the limited number of users on a network that was still in its infancy. AT&T’s speeds declined sharply between July and November 2013 but have stabilized over the past few months. Verizon and Sprint have declined slightly but remained relatively constant over the past year.
Since the T-Mobile rollout all networks have improved their coverage according to our metric, with the average user experiencing greater Time on LTE than a year ago across all networks. This improvement contrasts with speeds, where the story is less clear - possibly suggesting that LTE avialability is less negatively impacted by growth in the number of subscribers.