We review the construction and classification of three-family grand un
ified models within the framework of asymmetric orbifolds in perturbat
ive heterotic superstring. We give a detailed survey of all such model
s which is organized to aid analysis of their phenomenological propert
ies. We compute tree level superpotentials for these models. These sup
erpotentials are used to analyze the issues of proton stability (doubl
et-triplet splitting and R-parity-violating terms) and Yukawa mass mat
rices. To have agreement with phenomenological data all these models s
eem to require a certain degree of fine-tuning. We also analyze the po
ssible patterns of supersymmetry breaking in these models. We find tha
t the supersymmetry breaking scale comes out either too high to explai
n the electroweak hierarchy problem, or below the electroweak scale un
less some degree of fine-tuning is involved. Thus, none of the models
at hand seem to be phenomenologically flawless.