, the starch industry can receive up to 250,000 kg, or 10 truckloads, of potatoes per hour during peak season. The speed of the raw material flow, of course, depends on the number of conveyor belts used to handle the incoming raw material. It is unlikely that this amount can be handled by one single conveyor belt, and multiple instruments may be required to ensure good coverage. Brunt and co-workers made a prototype machine for prediction of starch and dry matter content in potatoes [29]. They briefly investigated the use of NIR for dry matter content prediction, but in the end traditional gravimetric prediction of the dry matter content was used. The authors stated that 10 samples per hour would suffice for their needs, each sample containing 3–5 kg. To sample one sample (5 kg) for each 10,000 kg is also common when using the gravimetric method for determining dry matter content. This equals 25 samples per hour (total of 125 kg of tubers), if calculating with the before mentioned amounts of potato. The conveyor belt in the current study was moving at 15 cm/s during the on-line measurements. Based on the number of spectra obtained from each potato tuber (approx. 15 spectra per tuber) it seems obvious that the speed can be increased. The prediction performance was largely maintained, provided that spectra used for dry matter content are samples from the entire length of the tuber. If increasing the speed to 25 or 30 cm/s the system would be able to measure nearly 300 individual tubers (with size of approximately 50 mm) per minute. This should be well within 10 samples per hour, and maybe also as much as 25 samples, allowing for different sampling regimes. Hence, the speed of the system demonstrated in the current study should have the capabilities to be incorporated in an on-line application. It is, however, expected that the current approach could be used at significantly elevated conveyor belt speed.