Advertising-based Measurement: A Platform of 7 Billion Mobile Devices

The most important step in an empirical computer scientist's research is gathering sufficient real-world data to validate a system. Unfortunately, it is also one of the most time-consuming and expensive tasks: placing measurement tools in remote networks or end-clients requires one to marshal resources from different administrative domains, devices, populations, and countries. Often such efforts culminate in a trace that is deficient in multiple ways: a small set of test subjects, a short time frame, missing ground truth for device IDs, networking environments lacking in diversity and geographic spread, or highly biased sampling. We present a method of addressing these challenges by leveraging the most open and globally accessible test and measurement platform: digital advertising. Digital advertising instantly provides a window into 7 billion devices spanning every county for an extremely low cost. We propose Advertising as a Platform (AaaP), an ad-based system to perform massive-scale mobile measurement studies. In contrast with measurements made by large media companies who own platforms, ad networks, and apps, we concentrate on the opportunities and challenges for researchers that are end-users of advertising systems. We evaluate a prototype system, discuss ethical guidelines, and demonstrate its use in four scenarios: IP2Geo databases, bandwidth measurement, energy management, and the identifiability of mobile users. We show the efficacy and ease-of-use of AaaP, and illuminate key challenges and the great promise of using AaaP to study a wide variety of mobile phenomena.