Sensor-Cloud Systems (SCS) are designed to link sensor networks and cloud applications, in order to gather and process information. While this is a hot topic for both academia and industry, with a large number of implementations, not much effort was put towards analysing the reliability of these systems. This article presents an experimental implementation of a system in the field of intelligent products and explores reliability improving techniques in five main areas of the SCS: network communication performance, auto recovery, local backup, automated software testing and system security. They all play an important role in determining the level of reliability of the novel Internet of Things (IoT) architecture framework for intelligent products presented in the article. A new formula is proposed for assessing the reliability of a SCS based on metrics from each of the five areas. Metrics used to assess the system reliability are presented, along with comparisons between operating with improvement techniques and without them. The results show that the reliability of the implemented SCS is improved considerably by implementing a deliberate reliability policy using an original five level tiered approach.