The transportation sector remains a major contributor to global carbon emissions. While private electric vehicles (EVs) are increasingly widespread, the electrification of heavy-duty vehicles (HDVs) that constitute fleets, including buses and long-haul trucks, is critical for achieving deep decarbonisation. However, the clustered and high-power charging demand of HDVs, particularly at depots, poses significant challenges for electricity distribution networks and charging station operators. Prior research has relied primarily on stochastic simulation models that, although valuable, often fail to capture localised variability, seasonal fluctuations, and operational factors for accurate grid infrastructure and charging schedule planning. This paper presents an empirical analysis of real-world anonymised charging data collected across diverse charging sites, including fleets of Rideshare vehicles, Shuttle buses, and Trucks operating in urban and semi-urban settings. By examining charge session characteristics and site-specific load profiles for various fleet types, we offer insights into their operational nuances, spatio-temporal variations and the predictability of load profile. This study shows that accounting for site type and usage, together with seasonal effects and charging behaviour, informs data-driven planning of resilient, cost-effective EV charging infrastructure.