A randomised complete block design (RCBD) has been extensively used in consumer testing; however, it is ineffective in preventing sample presentation biases. This study demonstrated a more efficient experimental design [i.e. split plot with repeated randomised complete block design (SPRRCBD)] to help minimise positional and first serving-order (FSO) biases in consumer tests. Results suggested that positional biases were strongly pronounced for the poor-quality sample. Comparing between RCBD and RCBD without FSO experiments, there were significant differences in overall liking (OL) scores of the same sample served at the same position. Hence, omitting the first sample score from data analysis was not recommended. The mean-squared error (MSE) of SPRRCBD was lower than RCBD and RCBD without FSO (2.28, 2.72 and 2.60, respectively), indicating a more powerful design to explain variations in mean hedonic differences. SPRRCBD extracted more explained variances, resulting in a decreased Type-II error in the model.