Cocoa bean winnowing has a function to separate cocoa nibs from shell after roasting process of dry bean. Nibs are further processed into fine cocoa liquor by refining process. The aim of this experiment was to evaluate working performance of a home-scale winnower to separate shell from nibs with minimum shell parchment content in cocoa nibs. This experiment was conducted in Postharvest Laboratory at the Indonesian Coffee and Cocoa Research Institute using roasted cocoa bean grade A according to standard of SNI 2323:2008/ Amd1:2010 with shell content of 15% originated from Forastero cocoa. Working performance of the home-scale winnower was evaluated based on shell parchment content in the output, its capacity, energy consumption and power transfer efficiency value by several air suction rates as variable. Data were analyzed using regression and variance analysis to evaluate the influence of the rate and to determine the optimum machine operation. Results of regression and variance analysis from winnowing experiment with air suction rate of 0.54 m/s; 0.63 m/s; 0.72 m/s and 0.90 m/s indicated that shell parchment content in cocoa nibs and power transfer efficiency value were affected by the rate. The optimum machine performance was obtained on 0.72 m/s of air suction rate with total winnowing capacity was 2.615 kg/hour, energy consumption of 132 Watt, power transfer efficiency value of 61.01% and shell parchment content was 1.06%. Shell parchment content in cocoa nibs was appropriate regarding to the SNI standard with maximum content of 1.75%.