The controversy that has surrounded herbivory studies in the last few decades prompted our investigation to establish the extent to which herbivore optimisation hypothesis or compensatory growth evidence is real. We used the traditional movable cage method to collect primary productivity data on herbage, functional groups and key individual grass species in various controlled large herbivore treatments in an east African savanna. The herbivore treatments in triplicate blocks included cattle, wild herbivores with and without mega herbivores and combinations of cattle and wild herbivores also with and without mega herbivores. The findings revealed that at herbage level, most grazed treatments (four out of five) had higher productivity than the ungrazed control and three showed grazing optimisation curve at sixth polynomial degree between monthly productivity and grazing intensity (1-g/ng). At functional group level forbs productivity was higher in the ungrazed control than in any of the grazed treatments while at individual grass species level Themeda triandra productivity was higher in all grazed treatments than in ungrazed control. We conclude against presence of herbivore optimisation hypothesis at herbage, functional group and species level because of lack of attributable grazing effect in grazed treatments that matches complex ecological effects in the ungrazed treatment.