By Celine Tran, Global Head of Operations, Axonix Ltd
If viewability is the new impression, and it has been for a while, location has become the new inventory currency.
Publishers and their supply partners know location data in a bid request upgrades inventory from economy class to premium and agencies seem happy to pay more for it on that basis, however concerns have been creeping in around the accuracy and authenticity of this information.
In programmatic, applications that pass GPS or other location-derived data via an SDK are important because they provide a rich mix of data points that help mobile traders better understand the ecosystem they are operating in, but these are meaningless if you don’t know for sure that a location signal is real or reliable.
At the heart of the matter lies the fact that a wide variation of exactitude exists between the location provided at media buying stage and the user’s GPS position when seeing the impression. Some of it is down to fraud but actually the majority is just down to the way different location technologies work and variances in how that data is cached by publishers and supply partners.
To understand the scale of the issue, recent research from think tank, Thinknear said 30% of all ‘hyperlocal’ location data in the programmatic ecosystem was accurate to within 100 m of a user’s actual real-time location – that means 70% is not.
The skill lies in being able to suss out all the red flags at scale and with programmatic mobile inventory revenues sitting at a projected $20.6bn for 2018 (BIA/Kelsey) there is no doubt it’s a high value sector for those that trade in it.
The incentive to clean these signals is clear and a number of ways to do this do exist ‘but the fact of the matter is you can have all the compression and machine learning algorithms you like to analyse bid requests and scrub mobile data, but if you’re working off an inaccurate base to start with – the whole exercise is pointless.
However, there is one identifier that can’t be simulated and which is immutable – a cell tower signal. If you can compare a phone’s GPS with pings from a mobile mast, you can place a user at a location with 100 pc confidence.
Axonix decided to see if that technology could be productised. Thanks to its relationship with global telecoms company Telefónica – it was able to develop a method which compares aggregated scores for app publisher GPS signals with the location of phones connected to any of Telefonica UK’s (O2) cell towers to verify location accuracy.
By combining millions of location comparisons, Axonix found it was able to index the sources of location data and score the quality of their location signals in this way that meant it could eliminate bad data and unreliable sources – resulting a type of Quality Vendor Index built on the best seed data available. Other location providers rely on probabilistic modelling and pattern recognition to distinguish between good and bad which risks inaccuracies by default whereas we have an actual truth set to verify against.
Everyone understands that accuracy lies at the heart of maximising benefits of programmatic trading but until a standard is set for collecting it, carrier grade verification is the way forward.