An autoscaling system for scaling resource instances in a cloud network includes an autoscaling application is stored in memory and executed by a processor. The autoscaling application is configured to provide an interface to define an autoscale policy including scale in rules and scale out rules for a plurality of different types of resource instances of a tenant. The autoscaling application is configured to receive capacity data corresponding to a first type of the plurality of different types of resource instances; calculate an estimated instance count for scaling in the first type based on the capacity data and scale in rules; calculate a projection factor based on an estimated instance count and a current instance count; generate adjusted capacity data based on current capacity data and the projection factor; compare the adjusted capacity data and the scale out rules; and selectively scale in the first type based on the comparison.