Insertion sort builds a sorted prefix by shifting larger prefix values to make room for each next value.
How to reason about it
Start by identifying the state the method maintains and the operation that changes it. Then trace one small input by hand. This separates the invariant that makes the method correct from implementation details such as arrays, pointers, counters, or protocol messages.
Worked example
For [2,4,3], inserting 3 shifts 4 one position and produces [2,3,4].
Write down the state before and after each step. For a larger input, repeat the same transition rather than guessing from the final output. That habit makes edge cases visible and gives you a concrete basis for testing.
Trade-offs and common mistakes
Its O(n²) worst case makes it unsuitable for large random inputs, although it performs well on small or nearly sorted data.'
Choose this technique when its assumptions match the problem. State those assumptions explicitly, test an empty or boundary input, and compare the cost of the operation you perform most often with the memory or setup cost it introduces.
Takeaway
Insertion Sort and Nearly Sorted Data becomes easier to use once its invariant, input assumptions, and failure cases are clear. A small worked example is a useful first test; measurements and proofs should follow when the system grows.