Much like the definition of goods vs service, you have it wrong. The best practice is to choose 100 sample size per 100k population. Most companies will use a lot less than that and try to focus more on cross-section. For instance, for population of 500k, with high confidence and error margin of 5 (which is the +-5 you see on all polls), a normal corporate sample size would be less than 350 people. In politics, however, more is better. The larger your sample size the higher your confidence, especially when you’re polling on concepts vs tangible objects. There were over 1500 votes in the cores poll. Even if only half were real, which I seriously doubt, that would still leave more than double the sample size required for 500k, which we all know is an inflated population number anyway. Thusly, I stand by those numbers and the interpretation of them.