An experiment was designed to investigate the optimum allocation of sampling farmers' raw tobacco for chemical analysis. A number of farmer lots was sampled from a number of villages in each district. Three samples per farmer were taken representing the three harvesting levels of leaves (bottom, middle, top), which were analysed for nicotine, total nitrogen and sugars. The estimated variance components for the three levels of sampling (farmers, villages and districts) were not uniform enough to lead to similar conclusions for all constituents. Farmer's within villages variance component was generally higher compared to villages and districts.