Preserving the privacy of users on web environments is a key requirement for analytics applications such as recommender systems, crowdsourced platforms, and web-scale data mining or machine learning applications. Yet, in spite of the existence of best effort anonymisation algorithms, incidents of privacy breaches remain a concern. In this chapter, we present existing mechanisms for transforming data to obtain usable and privacy preserving datasets both from the syntactic data transformation perspective. We focus on the vulnerabilities of these data transformation mechanisms to privacy breaches, in an effort to explain why and how privacy exploits occur. For simplicity, the examples we use to illustrate privacy vulnerability characterisations are based on a relational database model but can be extended to other data models or representations. As an added dimension and also for completeness we discuss data collection models and platforms, and the privacy implications of these models on data compositions in distributed environments such as the Internet. We wrap up the chapter with a discussion on the theoretical performance implications for data transformation algorithms, specifically in terms of generating anonymised data, in a bid to underline why most existing anonymisation algorithms can only guarantee sub-optimal privacy.