This paper addresses a set of methodological questions. First, it assesses the variation of the level of ANA in different non-hypothetical CA formats. Second, it explores whether asking respondents to report the attributes they ignored after each choice set or at the end of the choice task yield comparable results. Lastly, it explores the implications of taking into account ANA information on respondents’ willingness to pay and on the external predictive powers of the estimated parameters. To answer these research questions, three treatments were carried out, non-hypothetical CE (NHCE), non-hypothetical RCA (NHRCA) and nonhypothetical BWS (NHBWS). The results reveal that taking into account ANA information significantly improved the goodness-of-fit of the estimated models, especially when full ranking information is considered as NHRCA and NHBWS. In term of marginal WTP estimates, the results show neither of the two ANA approaches appears to be a clear winner. Also, our results show that taken into account ANA information not seem to improve significantly the predictive power of estimated parameters.