Given the absence of clear hints for new physics in the multiple channel-by-channel searches at the Large Hadron Collider, it is now more important than ever to adopt a different approach and try a more comprehensive strategy to see where beyond the Standard Model physics may hide. Instead of a typical top-down method, a more global, more model-independent search for new physics can be envisaged through an anomaly detection procedure based on a bottom-up approach. This chapter will focus on the “proto-modelling” project, whose specificity is to use simplified models to look for potential dispersed signals that may have been missed in the common analysis-by-analysis interpretations of the data. It uses SModelS and its database to find small excesses in the experimental results and to construct collections of simplified models maximally violating the Standard Model hypothesis while maintaining consistency with the ATLAS and CMS constraints. This chapter discusses improvements and novelties made to the “proto-modelling” machine (initially published as a proof-of-concept) and presents a first test run of this machine.