The breadth of a mobile network has traditionally been expressed in terms of geographical or population coverage, but we're evolving the standard to a metric that better represents how users experience a network connection: the proportion of time they have a connection.
Geographical coverage is perhaps the most traditional metric for looking at network rollouts. This refers to the proportion of land where a network can be accessed and is generally based on perdictive models. One of the biggest problems with this metric is that it doesn't take into account where people are. You might have a small proportion of the land with network access but it might be most of the population.
Generally this is expressed in terms of the percentage of households that have access to a network. This does take into account where people live but it fails on two counts. Firstly, it generally only accounts for network access to the doorstep but doesn't analyse if people have a connection indoors, or the effects of building matierals. Secondly, it doesn't take into account whether people have a connection all the time they are not at home, which is a very large part of 'mobile' usage.
OpenSignal's 'Availability' measures the proportion of time users have network access. By continually measuring whether users have a connection or not we are able to extend our assessment of networks to account for what happens when users are indoors and when they are moving around. We build up a holistic, user-centric measurmement of networks that expresses how users experience them.
The OpenSignal app takes a background reading from the device every 10-15 minutes. This reading includes many fields, such as the network technology - 2G, 3G or 4G, - and we use this to calculate the proportion of time that a user has access to signal. We also measure the proportion of time a user doesn't have access to any of these technologies.
The proportion of time on a given network type is then defined as the number of measurements on that network type, divided by the total number of measurements. We first calculate averages per device, and then average these results to give the operator-level time on LTE percentage. We use device averages to eliminate the effects of an unusual behavior profile from a device contributing a large amount of data.
Many people have LTE compatible phones, but do not have a plan which allows them to use LTE on their network. For this reason we look at all the data at the device level, and only calculate LTE Availability for a given device after we have verified that they have connected to LTE successfully at least once. This ensures that our LTE Availability metric only takes into account LTE compatible devices with an LTE compatible plan.