An attribute-based approach to modeling resid structure and reaction is described. Monte Carlo simulation of feedstock structure casts the modeling problem in molecular terms. This technique samples probability density functions (PDFs) for the attributes of the resid molecular structures to construct a representative molecular sample whose properties are compared against measured properties. Optimization methods are used to minimize the weighted sum of squares difference, and the final set of PDF parameters are the mathematical representation of resid structure. Subsequent reaction models can be based on the molecular attributes of the PDFs. The latter model, which is called the "Attribute Reaction Model" (ARM), provides a large reduction in the number of reaction equations and, therefore, solution time. The combined structural and reaction model provides a very good representation of laboratory data for four resids.