African savannas support an iconic fauna, but they are undergoing large-scale population declines and extinctions of large (>5 kg) mammals. Long-term, controlled, replicated experiments that explore the consequences of this defaunation (and its replacement with livestock) are rare. The Mpala Research Centre in Laikipia County, Kenya, hosts three such experiments, spanning two adjacent ecosystems and environmental gradients within them: the Kenya Long-Term Exclosure Experiment (KLEE; since 1995), the Glade Legacies and Defaunation Experiment (GLADE; since 1999), and the Ungulate Herbivory Under Rainfall Uncertainty experiment (UHURU; since 2008). Common themes unifying these experiments are (1) evidence of profound effects of large mammalian herbivores on herbaceous and woody plant communities; (2) competition and compensation across herbivore guilds, including rodents; and (3) trophic cascades and other indirect effects. We synthesize findings from the past two decades to highlight generalities and idiosyncrasies among these experiments, and highlight six lessons that we believe are pertinent for conservation. The removal of large mammalian herbivores has dramatic effects on the ecology of these ecosystems; their ability to rebound from these changes (after possible refaunation) remains unexplored.