Work out 23×47 in your head. As one problem it is awkward. Split 47 into 40+7: 23×40=920 and 23×7=161 are two easy products, and 920+161=1081. The split worked because the distributive law makes the two pieces independent: neither product needs the other's answer. Long division, partial fractions and finding an area piece by piece all rely on the same move.
A good decomposition has three properties. The pieces cover the whole problem, with nothing missing. They do not overlap, so nothing is counted twice. And each piece is easier than the original. Splitting an L-shaped floor into a 6×4 rectangle and a 3×2 rectangle passes all three tests, giving 24+6=30 m2. Splitting it into two overlapping rectangles fails the second test, and the shared patch gets counted twice.
Decomposition is not simplification. Rewriting x−1x2−1 as x+1 replaces one expression with an easier equivalent; there are no separate subproblems to solve. It is also different from case analysis, which splits by situation (if x≥0, if x<0). For any given x only one case applies, while every piece of a decomposition contributes to the final answer.
Think of it like this. Decomposition is how a removal company empties a house: the piano, the kitchen and the bookshelves each get their own crew and boxes, and no crew needs to know how the others are packing. The plan breaks down when the pieces are linked, like a bed frame bolted to the wall, which has to be handled together.
Core idea. Decomposition splits a hard problem into simpler, independent subproblems you can solve separately, then handle on their own.