Two of the experimental methods used to estimate willingness‐to‐pay (WTP) for a non‐market good, the Becker‐DeGroot‐Marschak (BDM) mechanism and the non‐hypothetical choice experiment (nHCE) often lead to significantly different WTP estimates, complicating the choice between the methods. In Zambia the same group of researchers used both techniques to evaluate WTP for orange maize, which provides more vitamin A than other varieties. This provided an opportunity to analyse the sources of the difference. In the BDM experiment, one group of respondents was provided with more training opportunities than the other, and made higher bids. Accounting for lexicographic behaviour in the nHCE reduced the estimated WTP. These two design factors together resulted in a decrease in the WTP difference for orange maize (1,279–632 ZMK) although the difference remains statistically significant. More training was also shown to eliminate the effects of different orders in which maize varieties were presented.