The important thing to consider is how precise you want your estimate to be of the 'correct value'' or 'truth' - that is, the amount or degree of variation that your sampled estimate would have of the entire population. The standard error is proportional to 1/square root of number of samples. Thus your expected standard deviation would be 1/root20 or ~ 22% for 20 samples. To bring this down to 10% would require 100 experiments/numbers/samples and to 1%, then 10000. Your estimate of the truth needs to be 'fit for purpose' or 'just good enough' and you haven't told us what your objective is.
See also: What is the minimum sample size for quota sampling? - ResearchGate. Available from: https://www.researchgate.net/post/what_is_the_minimum_sample_size_for_quota_sampling
In general, at least 30 samples are required for a good statistical estimate using normal distribution. Less sample will make your confidence interval wider.