The previous studies mostly explore the excess return on IPO. However, the risk is more important in IPO market since information asymmetry is more serious. Therefore, this study adopts market information, ownership structure and financial performance to explore the factors of failure in IPOs in Taiwan. Logic regression model and misclassification cost of Bortiz and Kennedy (1995) are used to compare the performances among models with different cut-off values. The results reveal that model with 0.15 cut-off value obtain relatively minimum misclassification cost. Finally, the significant variables that would affect the IPOs are CPA firms, financial leverage, EPS and market listing of categories. The accurate rate of classification in this prediction model is 82.9%.