Abstract: Electricity theft has long been a critical issue that jeopardizes the stable operation of electrical systems and incurs substantial economic loss. With the advent and widespread application of integrated electricity and line loss management information systems, significant strides have been made in the field of electricity theft detection. Utilizing data-driven methods, the large-scale user data collected enables more accurate identification of theft cases. This paper begins with an overview of the historical development of line loss business systems, followed by a review of current engineering methods for electricity theft detection. It then identifies shortcomings in existing methods and offers a future outlook, particularly focusing on the potential offered by the integrated 2.0 management system. The paper argues that the use of high-resolution data and multi-dimensional event information can significantly boost both the reliability and accuracy of theft detection, while also providing robust datasets for machine learning and deep learning training. Furthermore, it critically examines the limitations of currently utilized datasets and suggests avenues for future research. The paper also proposes enhancing theft detection techniques by expanding the scope to include non-high loss lines and districts, and through more nuanced user classification. These findings and recommendations aim to stimulate ongoing innovation and development in the electricity theft detection arena.