We have initiated a large project on identifying the requirements for developing a realistic and ground-up approach to simulating the formation of terrestrial planets in our solar system. As the first phase of this project, we present here the criteria that any model of planetesimal growth needs to fulfill in order to be self-consistent and produce reliable results. We demonstrate how these criteria emerge by revisiting runaway growth and carrying out a thorough analysis of its results. As our goal is to identify the pathway to a realistic model, we focus analysis on simulations where at the beginning, planetesimals are not artificially enlarged. We show how using uninflated planetesimals, as the first requirement for a realistic model, will result in a set of criteria naturally emerging from the evolution of the system. For instance, the growth times in simulations with uninflated planetesimals become comparable to the time of giant planet formation implying that any realistic simulation of planetesimal growth, in addition to using real-size planetesimals, needs to include the perturbation of the growing giant planets as well. Our analysis also points to a strong connection between the initial distribution of planetesimals and the final outcome. For instance, due to their natural expansion, initially isolated distributions, or a collection of initially isolated distributions, such as rings of planetesimals, do not produce reliable results. In a self-consistent and realistic model, where the initial conditions are supported by basic principles and do not include simplifying, ad hoc assumptions, the entire disk of planetesimals has to be simulated at once. We present the results of our analyses and discuss their implied criteria.