The fast expansion of Internet-based multimedia services like Netflix, Skype or YouTube - the so-called Over-The-Top Services (OTT) - have increased the quality requirements of these services from the point of view of the user. In this sense, it is important not only to ensure a good Quality of Service (QoS) level - technical metrics - but also a high level of Quality of Experience (QoE) - subjective, user centered metrics - in order to improve the satisfaction level of the customer, and therefore the revenues of the service provider. In this work we present how a formal model of an OTT service can be used in an online monitoring tool intended for the service provider, which aims to detect possible low-quality scenarios of a user. This information is used to take countermeasures in order to avoid the bad-quality configurations, aiming to maximize the final QoE of the user and, therefore, the revenues of the service provider. The proposed methodology consists in four main phases: model the service as an Extended Finite States Machine (EFSM), monitoring the state at the model of the user, calculate the user stories - and its QoE - based on the formal model and the user's actual state, and, use the previously calculated paths to make recommendations aiming to maximize the final QoE of the service