A class orders food for a party: pizzas at $12 and salads at $6. The rules are to spend at most $120 (12p+6s≤120), get at least 5 pizzas (p≥5), and get at least 4 salads (s≥4). The plan p=6, s=6 costs $108 and passes all three. The plan p=8, s=4 costs exactly $120 and also passes. The plan p=4, s=10 costs $108, under budget, but fails the pizza rule, so it is out.
On a graph, each condition is a shape. An equation like x+y=10 is a line; an inequality like y≤6 is a half-plane, everything on one side of a boundary line. The feasible set is where all the shapes overlap. With x>0, x+y=10 and y≤6, the overlap is the part of the line x+y=10 where x≥4: a ray starting at (4,6).
A system of equations alone, such as x+y=10 and x−y=2, usually pins down one point, here (6,4). Adding inequalities changes the job: instead of one exact answer you usually get a region of acceptable answers, or you find that the conditions clash and nothing works. The opposite error is to read the conditions as a menu where meeting any one is enough. In a constraint system they are joined by 'and', so every one must hold.
Think of it like this. Picture a stack of sieves, one per condition. Pour in every possible pair of values; each sieve lets through only the values that pass its test. Whatever reaches the bottom is the feasible set. If the holes in different sieves don't line up anywhere, nothing gets through, and the system has no solution.
Core idea. A constraint system requires the same values to satisfy every equation and inequality simultaneously.