Due to the diversification and fragmentation of working time arrangements, the organisation of working weeks now differ substantially from each other. To account for week-to-week variability in working time estimates in time diary research, it is important that respondents keep their time diaries on designated registration days. At the same time, this week-to-week variability might lead respondents to postpone their participation to convenient registration days. Research shows that, due to diverse time cycles, postponing participation leads to substantial bias. However, these findings pre-date online time diary methodologies. In online time diary studies, algorithms assign registration days and, therefore, postponement is solely done at the convenience of the respondent and no longer relies on the availability of the interviewer. Analyses of online time diary data from Flemish (Belgian) teachers (ndiary days = 59,969; nteachers = 8567) reveal that postponement depends on (the day in) the time cycle. This postponement is partially selective and thus leads to biased working time estimates. Some oversampling strategies are suggested to account for this possible bias, but a designated, consecutive 7-day time diary approach with postponement remains the recommended standard for collecting reliable working time estimates.