Happiness surveys based on self-reporting may generate unreliable data due to respondents' imperfect retrospection, vulnerability to context and arbitrariness in measuring happiness. To overcome these problems, this paper proposes to combine a happiness evaluation method developed by Ng (Soc Indic Res, 38:1-29, 1996) with the day reconstruction method developed by Kahneman et al. (Science 306:1776-1780, 2004a) to form a new happiness survey procedure. Distinguished from other surveys that ask respondents to rate their subjective wellbeing on a given scale, this happiness evaluation method provides detailed instructions to help each respondent determine a personal unit and use this unit to measure happiness in various occasions. While the day reconstruction method helps avoid the effects of imperfect retrospection and external disturbances, this new procedure can further reduce arbitrariness in self-reported measures and derive more accurate, intertemporally and interpersonally comparable happiness metrics. How to evaluate national wellbeing based on the data collected from such a survey is also demonstrated.