Understanding mobile network experience: What do OpenSignal’s metrics mean?

In this blog post, we’re taking a deep dive into the very core of OpenSignal: explaining the metrics we use, what they all mean and what their roles are in measuring the real-world mobile network experience as users see it.

How do we collect the data in the first place?

We collect and analyze more than 3 billion measurements every day, from more than 100 million smartphones across the world. We collect data every day of the week, at all hours and in all the places people live, work and travel: no simulations, no predictions, no idealized testing conditions. Our data comes from actual smartphone users and we report users’ actual network experience, whether they are indoors or out, bustling in a busy city or trekking in the countryside.

We collect the vast majority of our data via automated tests that run in the background, enabling us to report on users’ real-world mobile experience at the largest scale and frequency in the industry. These automated tests are run at random points in time and therefore represent the typical experience available to a user at any given moment.

Availability – Adding time to the equation

Our 4G availability metric shows the proportion of time users with a 4G device and subscription have an LTE connection. When we report an average 4G availability of 75%, that means our LTE users were, on average, connected to LTE services on their network 75% of the time.

Availability is not a measure of coverage or the geographic extent of a network.  It won’t tell you whether you are likely to get coverage if you plan to visit a far flung region that is off the beaten track. Instead, it measures what proportion of time people have a network connection, in the places they most commonly frequent, something often missed by traditional coverage metrics. Looking at when users have an LTE connection, rather than where, provides us with a more precise reflection of the true user experience.

We also keep track of the instances that leave mobile users most frustrated: when there is no signal to connect to at all. The most common dead zones users struggle with occur indoors. As most of our availability data is collected indoors (not surprisingly, since that’s where users spend most of their time), we’re particularly astute at detecting those areas of zero coverage.

Download Speed – The real speeds users get

When it comes to average download connection speeds we report publicly on three variations: 3G, 4G and overall speed. Measured in Mbps, our 3G and 4G speed metrics reflect the average download speeds our users’ experience on each type of network over the course of the data collection period.

Overall download speed represents the typical everyday speeds a user experiences across an operator’s mobile data networks. It is calculated as a weighted average of the individual 3G and 4G speeds based on the proportion of time typical users spend connected to each. If users cannot connect to an LTE network due to low availability, their overall speeds will also drop as they will spend more time on 3G connections (which can be up to 4 to 5 times slower than an LTE connection).

Upload Speed

Our upload speed metric measures the average upload speeds for each operator on LTE connections. Just as with our download speed metrics, 4G upload speed is measured in Mbps and reflects the average uplink speeds our users experience over the course of the data collection period.

Typically upload speeds are slower than download speeds, as current mobile broadband technologies focus resources on providing the best possible download speed for users consuming content on their devices. As mobile internet trends move away from downloading content to creating content and supporting  real-time communications services, upload speeds are becoming more vital and new technologies are emerging that boost upstream capacity.

Latency – Quantifying the lag

Latency refers to the delay users experience as data makes a round trip through the network. If the latency of your network is high, you’ll experience a lot more lag time. We measure our latency metrics in milliseconds for both 3G and 4G connections – the lower the latency value, the more responsive the network.

Both our download and latency speed tests are run on content delivery networks (CDN) that host the most popular internet destinations (not on dedicated testing servers where requests might follow a different route to normal traffic or indeed be prioritized in a way that isn’t representative), helping us to accurately understand typical user experience.

Time on Wifi

Our time-on-Wifi metric quantifies how often users spend connected to Wifi networks as opposed to cellular data connections. Time on Wifi doesn’t measure the amount of data consumed on Wifi. Rather it shows how often users’ devices are actively connected to a Wifi network.

Peak Speed – The fastest experienced speed

Peak speed is an average of the fastest speeds OpenSignal users experience on a network. For this metric, we disregard technical or congestion limitations and focus on the most optimized connections. This is different to the best-case speeds measured in idealized conditions that users themselves can never achieve.

To calculate peak speed, 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 user community. The average of that top 5% is our average peak speed metric. For a more practical grasp of peak speed, read our analysis on peak speeds in the U.K. and India.

Metrics in the pipeline

We are constantly evaluating new ways to measure users’ true mobile experience and frequently report on our findings as these metrics progress through the development stage. In time, we aim  to include these in our reports as standard. Here are a few:

Coverage Experience — The new way to look at coverage

Coverage experience is OpenSignal’s unique take on coverage. It contrasts with our availability metric in that instead of measuring when a user has a network signal, coverage experience is based on where a user has a signal. In short, coverage experience is the percentage of places where users’ devices have a network connection. We measure it as percentage, so if an operator has a 4G coverage experience score of 75% then our 4G users were able to connect to an LTE signal in 75% of all locations they visited during the designated test period.

Coverage experience is a far more reliable indicator of coverage than the traditional coverage metrics used by the industry. Traditional coverage metrics are based on mathematical models that try to predict how signals will propagate. This involves making a large number of assumptions about terrain, buildings, weather and many other factors all of which means the errors can compound and the end results may be quite inaccurate. Our coverage experience metric is based on measurements from real devices owned by millions of real users in all of the places they actually visit and spend their time.  We are not reporting on a simulated measure of the coverage people experience, as the industry has done for decades, but directly measuring it.

Video Experience

This metric measures the quality of experience for video streaming via a mobile network for 720p HD video based on an International Telecommunication Union (ITU)-vetted algorithm. Taking into account different factors including: picture quality, video loading time and stall rate, the video experience metric scores each operator on a scale from 0 to 100. Here’s how to interpret scores within that scale:

  • 83.3 – 100 Excellent: Generally video starts with almost no delay, plays continuously without interruption and renders at 720p HD resolution with optimal bitrate and high quality.
  • 66.7 – 83.3 Very Good: Video loads with only a small delay and typically plays without interruptions.
  • 50 – 66.7 Good: Typically video loads with a noticeable but not too long delay and is occasionally subject to short playback interruptions.
  • 33.3 – 50 Poor: Video loads slowly and prolonged playback interruptions are common.
  • 0 – 33.3 Bad: Video loads extremely slowly or stalls frequently for long periods of time.

With video being the single largest category of traffic carried on mobile networks and consumption expected to grow to keep pace with consumer demand, this is an extremely relevant metric. It is another key element in OpenSignal’s ongoing mission to measure the actual consumer experience on mobile networks.

We’re constantly working on new, better and more meaningful ways to measure users’ real-world mobile experience. Stay tuned to our blog for the latest updates and or sign up to receive our newsletter.  

Have a question or a comment? Leave a comment below.

Feel like checking the numbers for yourself? Download our OpenSignal or Meteor app. Both are, available on Android and iOS!

This entry was posted in Understanding signal and tagged , , , , , , , . Bookmark the permalink.

Leave a Reply