Read at the classic example of the Literary Digest poll and the Gallup poll trying to predict the results of the 1936 election between Alf Landon and Franklin Roosevelt.
The Literary Digest was wrong because of the method used which in reality was no statistic method at all:
The Literary Digest method was simple: to print up survey blanks and mail them to millions of households across the country. You simply had to fill in your choice for president, Landon or Roosevelt, and mail it back to the Digest.
By contrast, Gallup would conduct biweekly polls of a sample of perhaps 2,000 people — each one chosen, in the time-tested manner of market research, to represent a larger group, including all classes, races and regions.