DeviceAtlas comprises a fully transparent device description repository, with the data coming from device manufacturers, network operators, testing organizations, public domain sources, and DeviceAtlas customers. Registered users can view the source or provenance for every data point for every device. Browse the data here.
Some data is from public domain sources, such as the User-Agent Profile resource for devices. Some is licensed directly from device manufacturers, network operators and infrastructure and tools providers. Afilias web properties provide considerable levels of data due to their worldwide footprint and hence exposure to diverse devices. Finally, DeviceAtlas customers provide feedback on unrecognised user agents to ensure continuous improvement.
DeviceAtlas provides transparency in device detection, in that if a device is not recognised, a device ID is not returned. This allows accurate measurement of unrecognised traffic, which is a more robust and transparent mechanism than always returning a default device.
How the data model works
DeviceAtlas uses advanced algorithms to ensure data accuracy, with a patented method for accuracy assurance for multiple data sources. Each data source has an assigned ‘trust weighting’, which determines its influence over the result; for example a device manufacturer may assert that a device screen width is 300, but if over time other lower-ranked sources assert a different figure, the new result will be adopted once a threshold is passed. In this way there is continuous improvement of the data based on constant inputs from the DeviceAtlas user community, customers and partners.
DeviceAtlas data partners
DeviceAtlas has a growing list of contributing partners from different industry sectors, with a common interest in helping the mobile web developer to produce great sites and solutions. It's all about delivering the best possible results to the end user, no matter what they use to surf the web. To join us, contact us now.
How do you reconcile various sources of data?
Naturally we often get conflicting data from various sources regarding devices. In these cases, we use a weighted system to choose the most likely value for a given property. If everyone agrees the screen size of a device, then that's the one. If most people agree, then we take the most popular answer and adjust the value's confidence factor downwards slightly. Datapoints with conflicting sources are highlighted in yellow on the device pages.
How many devices and properties do you have?
At present DeviceAtlas identifies 34,000 devices from over 2,000 manufacturers, for which we provide 160 properties. New properties are added based on market inputs, and the number of devices is growing at a rate of 40 per day on average. You can see a list of all the properties and their descriptions here.
The device page shows many User-Agent strings, what do they mean?
Every browser, when visiting a website, sends a number of strings known as HTTP request headers. The user-agent string is one of these headers and is one of the best identifiers to recognize a device. DeviceAtlas uses this as part of its detection method, although all headers are factored in to ensure accuracy when third party browsers are used.
Where can I find a list of properties with a clear description?
We have a complete list of the properties we support available on the site.