A 2-3-4 tree is a balanced search tree where each node can store multiple keys. Because of that, it can keep its height under control as values are added. That makes searching and insertion efficient.
2-3-4 trees are useful because they give a clear example of how a tree can stay balanced while still supporting fast operations. They also connect to red-black trees, which represent similar balancing ideas in binary-tree form.
During insertion, you compare the new value with the keys in each node and follow the correct branch downward. If a node is already full, it can split as part of the insertion process. This helps keep the tree balanced.
Here is a page with a fuller explanation: UC Berkeley CS61B notes on 2-3-4 trees.