This study aims to compare and analyze the characteristics of risk reflected by potential risks based on crash modification factors and actual crash risk based on observed accident data at the homogeneous segment level for all general national highways in the Seoul metropolitan area. Conventional road safety evaluations mainly rely on observed crash data, such as crash frequency and rates; however, this approach has limitations in reliably estimating the risk for segments with low crash frequencies or no recorded crashes. To address this limitation, this study estimated the potential risk using crash modification factors that quantified the effects of roadway geometric design and environmental characteristics on crash occurrence, and the actual crash risk was calculated based on severe injury crashes over a recent three-year period. The analysis results indicate that the rank correlation between the potential and actual crash risks is very low, and the agreement rate between the two risk grades is also limited. In particular, segments with low traffic volumes tend to show a high potential risk owing to their structural vulnerability, whereas their actual crash risk remains relatively low. These findings suggest that potential both measures represent different dimensions of roadway safety and must be interpreted in a complementary manner. This study contributes to overcoming the limitations of single-indicator crash-based safety evaluations and provides an analytical basis for comprehensively understanding both the structural roadway and observed crash risks.