A tandem duplication denotes the process of inserting a copy of a segment of DNA adjacent to its original position. More formally, a tandem duplication can be thought of as an operation that converts a string $S = AXB$ into a string $T = AXXB.$ As they appear to be involved in genetic disorders, tandem duplications are widely studied in computational biology. Also, tandem duplication mechanisms have been recently studied in different contexts, from formal languages, to information theory, to error-correcting codes for DNA storage systems. The problem of determining the complexity of computing the tandem duplication distance between two given strings was proposed by [Leupold et al., 2004] and, very recently, it was shown to be NP-hard for the case of unbounded alphabets [Lafond et al., STACS2020]. In this paper, we significantly improve this result and show that the tandem duplication distance problem is NP-hard already for the case of strings over an alphabet of size $\leq 5.$ We also study some special classes of strings were it is possible to give linear time solutions to the existence problem: given strings $S$ and $T$ over the same alphabet, decide whether there exists a sequence of duplications converting $S$ into $T$. A polynomial time algorithm that solves the existence problem was only known for the case of the binary alphabet.