Does feedback usage lead to electricity savings? Analysis of goals for usage, feedback seeking, and consumption behavior
Findings from smart metering trials throughout the European Union in the last 5 years suggest that feedback on electricity consumption does not result in the expected increase in electricity savings of 10%. Instead, a smaller reduction in electricity consumption between 1.5 and 4% is typically recorded. In this paper, we aim to analyze knowledge and motivation gained from feedback usage and the resulting effects on behavior. For this purpose, we introduce a theoretical framework describing the process of feedback use in the Rubicon model (Heckhausen and Gollwitzer in Motivation and Emotion, 11, 101-120, 1987) and the transition from gained knowledge into motivation, and subsequent adaptation of consumption behavior, by applying the Feedback Intervention Theory (FIT, Kluger and DeNisi in Psychological Bulletin, 119(2), 254-284, 1996). Using log-files on usage behavior of a web-based feedback system, survey data and energy consumption data from a 1-year trial of smart metering systems and feedback with more than 600 consumers, we identify six patterns distinguishing how users interact with feedback systems (i.e., measured by the quantity of log-ins and clicks for different features) and are able to make inferences on their knowledge gains. Moreover, we examine whether any feedback usage pattern could be related to a stronger agreement with any goals. Finally, we compare the impact of these different usage patterns onto energy consumption in order to uncover which knowledge gains could be associated with energy conservation. Results show that feedback usage can be differentiated along several specific seeking strategies which are associated with individual goals. The only strategy of feedback usage found to be successful for saving electricity is actually the one showing the least evidence for long-term knowledge being gained. Possible explanations for these findings are discussed within the context of assumptions for the FIT.