Technological advancement can often act as a catalyst for scientific paradigm shifts. Today the ability to collect and process large amounts of data about individuals is arguably a paradigm-shift enabling technology in action. One manifestation of this technology within the sciences is the ability to study historically qualitative fields with a more granular quantitative lens than ever before. Despite the potential for this technology, wide-adoption is accompanied by some risks. In this thesis, I will present two case studies. The first, focuses on the impact of machine learning in a cheapest-wins motor insurance market by designing a competition-based data collection mechanism. Pricing models in the insurance industry are changing from statistical methods to machine learning. In this game, close to 2000 participants, acting as insurance companies, trained and submitted pricing models to compete for profit using real motor insurance policies --- with a roughly equal split between legacy and advanced models. With this trend towards machine learning in motion, preliminary analysis of the results suggest that future markets might realise cheaper prices for consumers. Additionally legacy models competing against modern algorithms, may experience a reduction in earning stability --- accelerating machine learning adoption. Overall, the results of this field experiment demonstrate the potential for digital competition-based studies of markets in the future. The second case studies the privacy risks of data collection technologies. Despite a large body of research in re-identification of anonymous data, the question remains: if a dataset was big enough, would records become anonymous by being "lost in the crowd"? Using 3 months of location data, we show that the risk of re-identification decreases slowly with dataset size. This risk is modelled and extrapolated to larger populations with 93% of people being uniquely identifiable using 4 points of auxiliary information among 60M people. These results show how the privacy of individuals is very unlikely to be preserved even in country-scale location datasets and that alternative paradigms of data sharing are still required.