Adding fractions goes one way: x−12−x+31=(x−1)(x+3)2(x+3)−(x−1)=(x−1)(x+3)x+7. Partial fractions runs it backward. You are given the combined fraction and want the simple pieces, because x−12 and x+31 each integrate to a logarithm, while the combined form matches no basic rule.
The denominator's factors set the shape of the pieces. Two different linear factors give x−1A+x+3B. Multiply through by the denominator to get x+7=A(x+3)+B(x−1), which must hold for every x. Choosing x=1 wipes out the B term and gives 8=4A, so A=2; choosing x=−3 gives 4=−4B, so B=−1. A repeated factor (x−a)2 needs two pieces, x−aA+(x−a)2B, and an irreducible quadratic such as x2+1 needs a numerator Ax+B; its piece integrates to a mix of a logarithm and an arctangent.
The method needs a proper fraction, one whose top has lower degree than its bottom. x−1x2+1 is not proper, and no constant A makes x−1A equal to it: for large x the original behaves like x, while x−1A shrinks to 0. Long division comes first, x−1x2+1=x+1+x−12, and only the proper remainder is decomposed.
Think of it like this. Partial fractions is unmixing paint: you are handed a colour and asked which pure colours, in what amounts, were stirred together to make it. The denominator's factors say which pure colours are possible, and solving for A and B finds the amounts. Unlike real paint, the unmixing is exact and has only one correct answer.
Core idea. Split a rational function into simpler fractions whose denominators are the original's factors.