Three data structures explain 80% of algorithm interview problems:
Hash map: O(1) lookup. First reach for it whenever you need to count, group, or look up by key.
Stack/queue: problems involving "process in order" or "undo" structure. Monotonic stack patterns solve a surprising range of range-maximum/minimum problems.
Two pointers / sliding window: linear-scan problems on sorted arrays or strings. Eliminates the nested-loop O(n²) that brute force usually produces.
The fourth: heap/priority queue, for "k-th largest" or "merge k sorted lists" families.
Most interview problems combine two of these. Recognition is the skill.