Simulation studies suggest that larger galaxies are formed through accretion of smaller galaxies. The process of accretion creates debris that contain only around 1% of the total mass but are spread over volume that is orders of magnitude larger and follow elegant and simple physics. Studying these structures will provide answers to fundamental problems such as (a) how often do the smaller galaxies merge into larger ones, and (b) what is the structure or locus of the debris. Traditionally astronomers have relied on predictions from simulations of how these debris structures form and behave. In the talk I will outline a method to identify the two substructures — shells and streams — from two-dimensional projected positions of a data cloud of stars. We first estimate the density of the two-dimensional position of the stars using a kernel density estimate (KDE) and then crucially use features of this KDE to identify shells from streams. The feature that distinguishes streams from shells is how the density of the data cloud behaves around the ‘ridge.’