Every few years, operators announce some big new network or network enhancement that promises to blow the doors off whatever our current mobile data speeds happen to be at the time. We see press releases proclaiming 300 Mbps, 600 Mbps, even 800 Mbps connections. But when these promised super networks finally arrive, the speeds that filter down to our smartphones and tablets are much slower. How does a 300 Mbps network suddenly produce a 25 Mbps connection?
That’s the tricky thing about cellular technologies. Mobile capacity is by definition shared capacity. Even if a mobile technology is theoretically capable of supporting speeds of 300 Mbps, that capacity has to be shared by every other user connected to a particular tower or cell in the network, so those 300 megabits get divided up among multiple devices. Even in the rare instance a single device is connected to a tower, it can’t access a network’s full theoretical speed due to any number of technical factors ranging from interference to distance — physics simply doesn’t cooperate. Finally, the performance of the wireless connection isn’t the only thing determining the speed of a mobile data link. All of the other components in between your phone’s processor and the server where your content resides plays a role in determining your internet connection speed.
Maybe you’ll never hit the advertised top speed your operator brags about, but that doesn’t mean there won’t be times where you’ll stumble onto an incredibly fast connection. Those connections are typically many times faster than the average speeds we measure on operators’ networks and give us a better indication of what the true technological capabilities of those networks are. At OpenSignal we’ve been experimenting with a new metric designed to measure speed when the network is performing at optimally. We call it average peak speed, and to kick things off we’ve calculated those optimal speeds for the four major LTE providers in India. You can see how they stock up in the chart below:
Before we dive into the data, though, let me explain how we define average peak speed. In order to calculate the metric, we only look at the fastest of our speed tests. Specifically, we only examine data from devices that have conducted multiple automated speed tests in a three-month period. We extract the fastest speed test from those devices and then toss out the bottom 95% of the results, leaving us with only the top 5% of the fastest speeds we’ve collected from our crowdsourced community. The average of that top 5% is our average peak speed metric. We’re not claiming this measurement is the topmost speed a device would ever experience on a particular operator’s network — operators could produce faster speed in the laboratory conditions or, in rarer cases, in the field. But we feel average peak speed is an accurate measure of a 4G connection under the most optimized conditions, and it gives an indication of what consumers might really experience when unfettered by congestion or technical limitations.
In the case of India, we calculated average peak speed from December 2016 to February 2017 so we could compare those numbers directly to the average 4G download scores we measured in our last India report. Just as Bharti Airtel won OpenSignal’s fastest 4G speed award in our report, we measured faster optimized speeds on Airtel’s networks than on the other three. Airtel’s average peak speed test was 56.6 Mbps, which is 5 times faster than its average 4G download test of 11.5 Mbps. When we look at Reliance Jio’s peak speed numbers, things really get interesting.
Even though Jio had the slowest average 4G speeds in our India report, in our peak speed comparison Jio’s average connection was second only to Airtel. We measured Jio’s average peak speed at 50 Mbps, which is nearly 13 times faster than its everyday 4G speed download speed of 3.9 Mbps. Meanwhile, both Vodafone and Idea peak speed measurements are about 4 times faster than their 4G measurements.
So why are Jio’s peaks so much higher than its average? Clearly Jio’s networks are capable of supporting some incredibly fast LTE connections when conditions are ideal. But what our data reveals is that those ideal conditions were far more rare for Jio subscribers than for customers of the other operators. The reason why is congestion. Jio has experienced phenomenal growth since launching its all-4G service last autumn, signing up more than 100 million subscribers in the space of few months, and for much of that period Jio has been offering those customers nearly unlimited access to mobile data. That kind of heavy usage is bound to tax any network, forcing users to vie against one another for bandwidth. Our data shows that Jio’s slow average 4G speeds aren’t a technical limitation, but rather a capacity bottleneck. As Jio adds more capacity — either through new spectrum or building more cell sites — or as Jio’s mobile data consumption levels drop, then its typical download speeds should increase.
Given the controversy in India over operator speeds since Jio first launched, our peak speed analysis explains quite a bit. Indian regulator TRAI has measured the fastest 4G speeds on Jio’s networks for the past six months through its MySpeed app. Meanwhile our testing along with measurements conducted by other independent testing groups like Ookla have found Jio’s average LTE speeds to be much slower. The discrepancy comes from TRAI’s methodology. TRAI is measuring operator performance under ideal conditions, while OpenSignal’s methodology tests the typical everyday experience of consumers. As our average peak speed metric shows, there’s a huge difference between Jio’s everyday speeds and its optimal speeds.
Of course, this is almost certainly a temporary condition. Jio has reined in the ultra-cheap unlimited 4G plans it first launched with. As Jio’s extraordinary data usage returns to more manageable levels, the big gap between its peak and average speeds should close. That means typical everyday connection speeds for Jio customers will increase as they’ll be able to access more often the full technical capabilities of Jio’s LTE network. We’re already seeing some evidence of this in our most recent data.
Our average peak speed metric is still in the experimental stage, so we haven’t included it yet in any of our full public reports. But the metric is part of OpenSignal’s ongoing effort to analyze crowdsourced data in new and interesting ways. We’ll tweak it as we go, and we’ll definitely be examining peak speeds in other countries shortly. In the meantime, let us know what you think and what other metrics you feel we should be looking at.