OBJECTIVES The EuroQol group observed problems related to the quality of valuation data collected using the 5L protocol. The objective of this study was to test if reintroducing a ranking task prior to TTO could help to improve the quality of the data. METHODS Respondents were assigned to three different study branches in Spain, while just two branches were used in Japan. The control group (Arm 1, both countries) was the 5L protocol. The second branch (Arm 2, just Spain) was the EQ-VT preceded by a ranking task, in which the respondents were asked to rank the target health states using physical cards. The last branch (Arm 3, both countries) was with ranking prior to the EQ-VT, but unlike the second branch, the ranked states remained visible through the Composite TTO (C-TTO) tasks, and the C-TTO states order were determined by the rank order. A sample of respondents was randomly selected from Canary Islands and Tokyo population. All interviews were performed face to face. We compared the number of C-TTO-based inconsistencies. RESULTS The final sample size was 932 respondents (141 and 191 for branches 1 and 3 respectively in Japan and 196, 205 and 199 for branches 1, 2 and 3 respectively in Spain). The percentages of inconsistencies on C-TTO responses (state 55555 not valued as the worst state) were 13.27%, 10.73% and 13.57% for branches 1, 2, 3 respectively in Spain, and 20.6% and 10.5% for branches 1 and 3 in Japan. Grouping the data, the inconsistencies were 16.3%; 10.7% and 12.1%. However, none of the observed differences were statistically significant. CONCLUSIONS Differences in proportions observed in the Spanish data are smaller than in Japanese data. Given that Japanese team did all control arm interviews prior to ranking task interviews could play in favor of ranking arm (learning effects).