The Electronic Nose (E-Nose) response has been optimized by an experimental design aimed at defining a standard method to analyze blend tobacco. A possible application of this apparatus is to use it as an "on line" quality control tool in manufacturing plants. This will enable us to present a correlation with sensory evaluation. We have carried out a follow-up of a US blend from plants during several weeks and we have analyzed head space of cut tobacco by E-Nose in order to control Volatile Organic Compounds (VOC) fingerprints of finished blends. In this purpose we suggested to make a survey chart based on euclidian distances calculated between each sampling and the average fingerprint of all samplings. According to the survey chart, we detected outliers which were not the object of customers' complaints and which did not present chemical differences. We decided to organise a blind tasting session by an expert panel. The outliers and a reference have been tasted in monadic and the sensory difference between the samples has been checked. In conclusion we can confirm that there are real differences between outliers and reference but it is undetectable by consumers. E-Nose is more sensitive than a human sense of smell. So as to use an E-Nose in order to decide if the differences between manufactured samples are significant or not, it is necessary to define an upper control limit that would be a "consumers' threshold limit".