Every finite abelian group is isomorphic to a direct product of cyclic groups of prime power order: G ≅ Z/p₁^(a₁) Z × Z/p₂^(a₂) Z × ⋯. This decomposition is essentially unique and completely determines the group.
From your work with direct products of groups, you know how to build new groups by combining old ones: Z/2Z × Z/3Z gives a group of order 6, Z/2Z × Z/2Z gives a group of order 4, and so on. The Classification Theorem turns this around — it says that *every* finite abelian group is built this way, from cyclic pieces of prime power order. You don't need to guess the structure; the theorem tells you exactly what the pieces must be.
The elementary divisors form of the theorem is the most concrete. To classify a group of order n, factor n into prime powers: if n = p₁^(a₁) · p₂^(a₂) · ⋯, then each prime contributes a direct product of cyclic groups whose orders are prime powers for that prime. For example, groups of order 12 = 4 · 3 = 2² · 3 come in two flavors: Z/4Z × Z/3Z ≅ Z/12Z (cyclic), or Z/2Z × Z/2Z × Z/3Z ≅ Z/2Z × Z/6Z. These are the only two abelian groups of order 12, up to isomorphism — there are no others. The invariant factors form gives an equivalent description using a chain of divisibility: G ≅ Z/d₁Z × Z/d₂Z × ⋯ where d₁ | d₂ | ⋯. The cyclic group Z/nZ corresponds to the single invariant factor n; non-cyclic groups have more than one factor.
Uniqueness is the theorem's muscle. Without it, classification would be a list of possibilities with no guarantee of completeness. Uniqueness says: if two such products are isomorphic, they have exactly the same set of prime-power cyclic factors (counted with multiplicity). This gives a complete, non-redundant catalog — to determine whether two finite abelian groups are isomorphic, compute their elementary divisors and compare the lists. Identical list? Same group. Different list? Different groups.
The proof strategy combines two ideas you should now find familiar. First, every finite abelian group decomposes into its p-primary components — the subsets of elements whose orders are powers of a fixed prime p. These components are themselves groups, and the full group is their direct product (one per prime dividing the group's order). Second, each p-primary abelian group decomposes into a product of cyclic p-power groups. This second step is the harder one; it uses the fact that in an abelian group, taking quotients and finding complements behaves much more predictably than in non-abelian groups. The result is a complete structural classification — a theorem with no analogue for non-abelian groups, where the story is far more complicated.