If I recall, the Reader's Digest polls was inaccurate because unlike Gallup they would received letters from their readers who would tell which candidate they would vote. The readers who sent in were wealthier and much better off than most in 1936 and thus were less representative of the entire electorate.
You would have to change the method of polling to what Gallup did in 1936, but if you are asking how one could make Landon president with the same result as what the Reader's Digest was getting then I can't help you there as my knowledge of this period is not high.
