Abstract: Shelf-life prediction model for power bar with color as the deterioration index was established based on Accelerated shelf- life testing ( ASLT) . Five factors related to shelf life prediction were evaluated, including the number of replicates, the number of fitted points, time interval, temperature range and the number of temperature points.The results showed that the number of temperature points and the temperature range were more important than that of other factors. The number of fitted points and time interval were less important factors for prediction error, and the affect of the number of replicates on the accuracy of shelf life prediction could be negligible. The relative predictive error could be below 10% by accurate experimental design.