Volcanic lightning shows considerable promise as a monitoring and research tool to characterize explosive eruptions. Its key strengths are rapid and remote detection, because the radio signals produced by lightning can propagate thousands of km at the speed of light. Despite these tantalizing properties, the scientific work on volcanic lightning has only recently started gaining momentum. Much more is needed to understand what lightning reveals about the evolution of an eruption in near-real time. Here we examine the timing and energy release of lightning generated by the eruption of Kelud volcano in Indonesia on 13 February 2014, as detected by the World Wide Lightning Location Network (WWLLN). The eruption column reached at least 26 km above sea level, representing the highest plume since the advent of global lightning networks in the last decade. Therefore, it provides valuable constraints on the electrification of end-member, sustained Plinian columns. We investigate the lightning in context with satellite images, photographs, and other published studies. Results show that the earliest satellite-detected activity was a thermal anomaly at ~15:46 UTC, corresponding to a directed blast at the onset of eruption (and only a few lightning strokes). Following a brief pause, the eruption produced a sustained column and umbrella cloud that spread outward into the tropical stratosphere. Rates of umbrella expansion provide an average mass eruption rate (MER) in the range of 8 × 10⁷–1 × 10⁸ kg s⁻¹. A more nuanced picture emerges from the time-varying MERs (determined between each satellite pass), which show rapid intensification during the first hour of eruption, followed by constant MER for about an hour, and waning toward the end (after ~17:50 UTC). At this stage, decreasing flux into the umbrella cloud coincides with column instability and formation of pyroclastic density currents, as recorded by photos from the ground ~17:45 UTC. We infer that some of the erupted mass partitioned into ground-hugging currents, leading to a lower apparent MER. Interestingly, there is not a 1:1 correlation between lightning intensity and MER over the course of eruption. Stroke rates increase sharply within the first 30–40 min (during rapid intensification of the plume), and then drop below 2 strokes per min once the MER remains constant. This suggests that electrification was controlled by the rate of increase in MER—in other words, the acceleration of particles out of the vent. We also show that lightning stroke-rates and energies are greatest within 50 km of the vent, even when the ash cloud extends >200 km downwind, indicating that lightning was focused in the regions of highest particle concentration and turbulence. Overall, we conclude that abrupt changes in lightning rates are clearly linked to changes in eruption behavior, and that rapid detection could aid monitoring efforts to characterize eruption rates or styles.