Willingness-to-pay (WTP) studies for traditional food products are plausibly affected by unobserved decisions and strategic collusion between the experimenter and respondents. Similarly, WTP estimates in developing countries using a one-time survey might be inconsistent, as the acceptance of new products likely varies with exposure to product attributes. We use repeated experimentation, where subjects are randomised twice on treatments, to reduce hypothetical bias and account for dynamic convergences of consumers’ preferences. We rely on longitudinal variation in treatments, which allows subjects’ characteristics and setting to have little influence on WTP estimates. These experimental designs evaluated consumers’ preferences for cakes from high-quality cassava flour (HQCF) and wheat flour mixtures in Nigeria.