Abstract The issue "to round or not to round" a linear programming solution when fractional answers have no meaning has long been a subject of concern to management. Professor Alfred Rappaport has shown that rounding can provide better answers than indiscriminate use of integer programming, particularly in the uncertain marketing and cost environment faced by many managerial decision makers. This note presents the other side of the coin, showing that even under uncertainty rounding can be dangerous, and the whole number solutions provided by integer programming (intelligently applied) are generally to be preferred. A simple numerical example is given of a problem whose linear programming solution can be rounded in more than two thousand ways, none of them giving a feasible, let alone optimal, integer programming solution.
Fred Glover (Tue,) studied this question.