Treatment vs. Control group Daily Energy Use
The feedback technology in this work, the Google Powermeter, resembled the technologies being deployed by several utilities in the United States and elsewhere. The households of 1,743 Google employees that participated in this study were recruited in collaboration with Google, both in their California offices and with several offices across the continental United States. Participants installed power monitoring devices, completed a survey and were randomly assigned to no-feedback (untreated control) or feedback (treatment) conditions.
Electricity used Oct 29-Oct 30, compared
to others and past usage.
Among the findings:
Initially, participants in the treatment group cut electricity use 5.7 percent more than those in the control group.
However, significant reductions were short lived. By week four all statistically significant reductions ended.
The largest reductions were observed initially at all times of the day. As time passed, evening reductions faded but morning reductions were sustained for eight weeks. However, the return to baseline in other day and evening periods cancelled out statistical significance in overall reductions.
- Stanford Energy and Behavior Feedback Survey (Pre-Feedback)
- Google Surge Post-Survey
- A Survey for Evaluating the Effect of Feedback Interventions on Energy Behaviors (a description of the survey tool)
Publications and Presentations
Real-Time Feedback and Electricity Consumption: A Field Experiment Assessing the Potential for Savings and Persistence
The Energy Journal, 34, 1
Houde, S., Sudarshan, A., Todd, A., Armel, C., Flora, J. (2013)
Future work should attempt to address some of the inherent challenges for robust trials of feedback and energy consumption: the heterogeneity of electricity consumption, the relatively low predictability of levels of that consumption, size of samples needed for detection of effects in the face of large heterogeneity. Also, data should be collected over periods of one year or more to assess seasonal and weather effects, and to achieve persistence through interventions of this type. Future work should also expand on analysis techniques to improve learning from trials, as well as to offer utilities and government entities trusted approaches to quantifying the impacts of behaviorally oriented programs.