Improving industrial land use efficiency is of great significance for sustainable industrial development. This paper measures the industrial land use efficiency (ILUE) in China, and explores its spatial-temporal characteristics and influencing factors by using a stochastic frontier analysis (SFA) method. A translog SFA model that characterizes the relationship between industrial land productivity and land input level is established to estimate the efficiency. The estimation results of the SFA model and related tests both show that the stochastic frontier method for ILUE measurement is suitable, and the contribution rate of technical inefficiency term to the mixed error reaches 66.9%. ILUE levels in China show an increasing trend from 2004 to 2017. The average efficiency in the east is the highest. ILUE in the west gradually surpassed that in the northeast in recent years. Reduction of the inter-group difference greatly contributed to the decrease in overall difference, but it is still the main source of overall difference. Results of the inefficiency function show that enterprise ownership structure produces negative impact on ILUE. The specialization of high-tech manufacturing industry, urbanization level and transportation condition have positive influence on ILUE, while the specialization of the chemical and rubber industry and the local financial pressure exert negative effects. Based on these, some policy recommendations for ILUE improvement are further proposed.