This paper addresses the evaluation of an energy efficiency program for low-income households. The so-called ‘éco-social' program was launched by the local utility of Geneva-Switzerland at the end of 2009. By the end of 2013 it had reached more than 5000 households through eight campaigns and attained its initial objective of 2 GWh/year of electricity savings before the end of the program. The aim of éco-social is to favor the implementation of energy efficient technologies in low-income households. All of the inefficient lighting are replaced by energy efficient lighting devices, outlet power strips with external switch are installed, electric kettle boilers are offered and rebate vouchers are given to replace existing refrigerators by energy efficient ones. Different types of data collection and methods are used to estimate the savings generated by the program: simple engineering ex-ante estimates (based on the information collected on the replaced and installed devices; enhanced ex-post engineering estimates (based on measurement of the energy consumption before and after the implementation). The ex-post methods were intended to calibrate the ex-ante methods that are used at present time to evaluate the savings during a given campaign. The methodologies and main results obtained are detailed in this paper. In addition, three topics are discussed: problems with the reference group for one of the saving estimation methods, the evolution of the stock of inefficient light bulbs in households, and the development of a software tool to estimate the savings. One of the methods used to estimate the savings is based on the comparison of participant and non-participant households. Since the first campaign, the program had a great success concerning the willingness of targeted households to participate. This complicated the task of evaluators who intended to use non-participants as a reference group and had to slightly change the way saving estimations were made. Data collected during the campaigns allow to follow the evolution of the stock of inefficient light bulbs in households. While a shift from incandescent light bulbs to compact fluorescent light bulbs (CFLs) was expected after the ban of incandescent light bulbs, it is rather observed a shift towards halogen bulbs. Statistics of the evolution of the average power of inefficient light bulbs, which are replaced during the campaigns, are also given in the paper. Using the calibrated ex-ante methodology presented in this paper, the program operator is able to estimate the savings generated by each campaign with relatively good accuracy. This has been introduced in the software tool that is used by program implementers to manage the campaigns and allows to have a picture of the energy savings generated almost in real time. The experience gained by the program and implementers allowed to redesign a new program for 2014-2015 with a more ambitious objective than the previous one. A similar objective (2 GWh/year) is targeted within a shorter time frame.