KIF1A-associated neurological disorder (KAND) encompasses a group of rare neurodegenerative conditions caused by variants in KIF1A, a gene that encodes an anterograde neuronal microtubule (MT) motor protein. Here we characterize the natural history of KAND in 117 individuals using a combination of caregiver or self-reported medical history, a standardized measure of adaptive behavior, clinical records, and neuropathology. We developed a heuristic severity score using a weighted sum of common symptoms to assess disease severity. Focusing on 100 individuals, we compared the average clinical severity score for each variant with in silico predictions of deleteriousness and location in the protein. We found increased severity is strongly associated with variants occurring in protein regions involved with ATP and MT binding: the P loop, switch I, and switch II. For a subset of variants, we generated recombinant proteins, which we used to assess transport in vivo by assessing neurite tip accumulation and to assess MT binding, motor velocity, and processivity using total internal reflection fluorescence microscopy. We find all modeled variants result in defects in protein transport, and we describe three classes of protein dysfunction: reduced MT binding, reduced velocity and processivity, and increased non-motile rigor MT binding. The rigor phenotype is consistently associated with the most severe clinical phenotype, while reduced MT binding is associated with milder clinical phenotypes. Our findings suggest the clinical phenotypic heterogeneity in KAND likely reflects and parallels diverse molecular phenotypes. We propose a different way to describe KAND subtypes to better capture the breadth of disease severity.