There is a gap between run-time service behaviours and the contracted quality expectations with the customers that is due to the informal nature of service level agreements. We explain how to bridge the gap by formalizing service level agreements with metric functions. We therefore discuss an end-to-end analysis flow that can either statically verify if a service code complies with a metric function or use run-time monitoring systems to report possible misbehaviours. In both cases, our approach provides a feedback loop to fix and improve the metrics and eventually the resource configurations of the service itself.