Satellite-based estimates of rainfall are frequently used to complement scarce networks of gauges. Understanding uncertainties is an important step, but is often hindered by a lack of validation data or misrepresented by spatial scale-related uncertainties, which are especially important in spatially-variable regions such as mountains. This study evaluates the Integrated Multi-satellitE Retrievals for GPM (IMERG) V05B 30-minute estimates for all three runs (Early, Late, Final) over the high tropical Andes. A unique dataset containing 15 rain gauges located within one IMERG grid at elevations ranging from 3,800 to 4,600 metres, provides a first evaluation opportunity in this topographical context. The evaluation was based on categorical, statistical and graphical methods. Error dependencies on precipitation characteristics and data source of the IMERG estimate were investigated. We show that IMERG severely underdetects precipitation events, thus underestimating precipitation depths. Poor detection is partially attributable to the low-intensity nature of precipitation over the region. However, tracing the error to the data source highlights limitations in passive microwave retrievals over the full range of intensities. No IMERG run has best overall performance, emphasising that run suitability is application specific. The impact of gauge density on performance metrics was also evaluated, and showed that sub-daily IMERG accuracy is overestimated by sparse networks. A minimum of six gauges was required at the 30-minute increment so that performance metrics are within 0.1 points of their true scores. We provide the first comprehensive assessment of 30-minute IMERG in a mountainous setting, highlighting the importance of high-density networks for accurate sub-daily evaluations.