One of the main cocoa producing areas in Indonesia is the Central Sulawesi region. Cocoa farmers have been given extension and training about agricultural technology such as pruning, fertilizing and side grafting. Nevertheless, the cocoa productivity is still low. This case has been questioned by authorities; so this research aims to measure and analyze the factors that have affected the variation of total factor productivity (TFP) in cocoa farming in the Central Sulawesi. The study selected a total of 424 cocoa farmers using a simple random sampling technique. The model of stochastic frontier function of translog forms was used to anticipate factors affecting cocoa production. The Cobb-Douglas production function was used in TFP variation. The results showed that TFP of cocoa farming varied from a minimum of 0.371 to a maximum of 1.407 with an average of 0.803. Technical efficiency, education, extension access, cocoa farming experience, sanitation, and fermented cocoa beans affected TFP variation significantly. These six factors appeared to be the main determinant factors of TFP variation. However, technical efficiency was the most important factor, so policies should aim at mobilizing resources toward expanding education and extension. In addition, extension services should provide information to cocoa farmers on using inputs efficiently.