I am afraid that there is no rule. The subdivision of species into finer taxonomic categories is often due to slight morphological differences, but it also depends on the taxonomic tradition of that particular taxon (it might even change among different families of the same Phylum). Subjectivity is a tricky question in taxonomy and several authors even criticize the validity of Linnean ranking. That is, if splitter taxonomists prevail you usually have a pletora of subspecies (slight morphological differences are overestimated), the opposite with lumper taxonomists (slight differences are ignored) . Molecular taxonomy might help (perhaps), but I am convinced that a general rule will be hardly found.
As far as the subjectivity of Linnean ranking you can easily check it on the internet.
The subdivision of species into subspecies is due to slight morphological differences together with an own geographical area. Nevertheless this second is avoided and as a result there are a lot of subspecies bad defined.
Many species are usually subdivided into infraspecific taxa, i.e. varieties and subspecies, based on minor differences in their morphological characters. However, the geographical separation is necessary for description of subspecies within a given plant species.
I think, in the taxonomical point of view, the main criterion for designation of sub-species or varieties would be morphological distinction. However with the presence of cryptic and pseudocryptic species especially in phytoplankton, one can hardly delineate the closely related species to species-level with only morphological observations.
With the well studied of nuclear encoded second internal transcribe spacer (ITS2), people had now used this molecular marker to serve as a tool in delimit the crypticity. ITS2 is a fast evolving region and have a very conserved secondary structures throughout the eukaryotes with four helices in the structures. ITS2 structures will help provide a better sequence-structure alignment for further phylogenetic analysis. Besides, according to a few researchers especially Coleman, there are sexually incompatible if there is presence of compensatory base change (CBC) in the secondary structures, sometimes even hemi-CBC to differentiate two closely related species. ITS had been proved useful in this matter not only in phytoplankton but also higher plants, insects and etc.
The use of subspecies rank in the classification is dependent on the species concept that one would like to rely on. The Phylogenetic Species Concept denies subspecies rank at all, declaring each variation that is distinct enough from the nominal species a new one. That is very popular currently, especially when using the molecular traits (e.g., mtDNA), regardless is that really justified. If you would like to stick to the Evolutionary Species Concept, I suggest you to explore the level of divergence (i.e., difference) in the population in concern: if it exceeds the average distance value for the trait used in the series of populations, it maybe could be recognized as a distinct subspecies. The useful precondition for that is also the limitation in gene flow between the new subspecies and the nominal (or other) one(s).