When searching for confirming hypothesis on a social domain, scholars regularly form samples to be used for the research. Analogously, for instance, market researchers do. Social networks constitute a huge source of information for these investigations, and it is often used to collect these data. When the collection of the data is made rigorously, an important effort has to be put on the selection of the sample, that should respect criteria of correspondence to the ditribution of the population, balance, and minimality of sample classes, in order to guarantee good inferential statistics evaluation parameters. In this paper we investigate a systematic way of providing samples on social networks that takes in consdieration the above mentioned criteria shile forming the sample. We prove that an affordable and reliable method to construct the basic instrument for the above method (the random table) can de provided, show its computational properties and devise an experimental protocol (the viral experiment protocol) that makes direct usage of the method mentioned above.