Hld is useful when u are given some kind of non-invertible operation queries with updates. For example, consider the problem witht the following 2 queries:

1. Update the value of node X to Y

2. Say the GCD of the path u...v

Can this be done in some other way ?

Apart from update queries, ofc:

HLD: O(N) preprocessing, O(logN) LCA query, O(N) memory.
Binary lifting: O(NlogN) preprocessing, O(logN) LCA query, O(NlogN) memory.
Sparse table on DFS order: O(NlogN) preprocessing, O(1) LCA query, O(NlogN) memory.
Segment tree on DFS order: O(N) preprocessing, O(logN) LCA query, O(N) memory.

It's easy to see that HLD beats binary lifting by any means (except for simplicity which is actually quite important in CP), sparse table has a better query time but worse preprocessing time and memory complexity while segment tree gives you the same complexities. Not sure if the constant factor for querying segment tree is bigger than the one for LCA query in HLD but I guess so.