Internet of Things (IoT) deployments are becoming a de facto solution for monitoring all types of devices, systems, and infrastructures in various fields. Furthermore, IoT is considered as a key element for implementing digital twins. The architecture of IoT deployments is usually composed of three entities: end nodes with embedded sensors able for data capturing, an intermediate element (gateway) that receives data from end nodes and forwards it to a cloud, and the cloud, which can be a (virtual) server responsible for processing the received data, creating information and knowledge, offering a proper visualization, and in some cases, taking actions as a response. Two essential tasks in IoT systems are to measure the level of quality of the offered service and to predict quality drops during the service operation. Here we present a tool called QUITO (QUality of IoT Operation) that measures the level of quality of the offered IoT service in four different dimensions, called Quality of Data, Quality of Information, Quality of user Experience, and Quality Cost. Then the performance of the service is predicted using Deep Learning, which enables taking countermeasures to avoid quality drops before it can happen. We implemented the tool and tested it in simulated scenarios, assuming the use of LoRa and LoRaWAN as one of the most common protocols within the Low Power Wide Area Network technology. Results show that QUITO can monitor and forecast the performance of IoT systems with a granularity level not currently present in other proposals.