For example, let's say (for the sake of illustration) that you have made the following guesses:
DINGO - 2
GRAFT - 0
DRIFT - 1
Do the logic and then write, off to the side, "DI=1" and "NO=1". Many players would, at this point, take one guess (e.g., DRAFT) to get the D or I, and another guess (e.g., GRANT) to get the N or O. However, what if you test one letter from each pair (e.g., FAINT)? If it scores zero or two, then you have pinned down both secret letters (DO or IN) at once! And if it scores one, although that's ambiguous, you will have logically linked "DI=1" with "NO=1" via "IN=1": Testing any one of those letters will now tell you the state of all four. (e.g., if DRAFT scores 1, then there's a D, so there's no I, so there's an N, so there's no O.)
This strategy, in the abstract, can be extended to larger amounts of information. Instead of carefully narrowing down your letters, guess words that have a wide range of possible scores. At either end of the range, you'll gain a great deal of information. And even in the middle, you'll have information that, while ambiguous now, will be useful later on.
For example, in one game, I made the following guesses:
QUACK - 1
VIXEN - 1
GLYPH - 2
TOMBS - 1
Having hit all five letters in the secret word, I selected the most common letter from each guess (the two most common from GLYPH), and arranged them into my next guess, LATHE. (These were the most common letters by written English usage, not by Jotto frequency. Oops.) I figured that was likely to score two or three, and would provide a very useful link between my previous guesses. Once I narrowed down the results from LATHE, I very likely would have confirmed the secret letters from two of my guesses, and have eliminated some letters from the other two. As it turned out, though, LATHE scored four, so I won in short order.