An example: Let's imagine we would like to check how many Flora tooth paste users are found in a specific population. We will be satisfied with 95% assurance that the result which we get from a survey varies not more than 4% from the true number in this population. If we predict there are 50% (0.5) Flora tooth paste users in the population, the sample size should be 600 persons (assuming an extremely large population). If the share of Flora tooth paste users in the population is more or less than 50%, the estimation error will be less for the same sample size. Usually we assume a value of 50%, which gives us 95% assurance that, for other percentages, the estimation error will be less than 4% (i.e., the so-called maximum error for estimation).