Sources of variation in nutrient intake have been examined for Western diets, but little is known about the sources of variation and their differences by age and sex among Koreans. We examined sources of variation in nutrient intake and calculated the number of days needed to estimate usual intake using 12 d of dietary records (DR). To this end, four 3 d DR including two weekdays and one weekend day were collected throughout four seasons of 1 year from 178 male and 236 female adults aged 20–65 years residing in Seoul, Korea. The sources of variation were estimated using the random-effects model, and the variation ratio (within-individual:between-individual) was calculated to determine a desirable number of days. Variations attributable to the day of the week, recording sequence and seasonality were generally small, although the degree of variation differed by sex and age (20–45 years and 46–65 years). The correlation coefficient between the true intake and the observed intake (r) increased with additional DR days, reaching 0·7 at 3–4 d and 0·8 at 6–7 d. However, the degree of increase became attenuated with additional days: r increased by 13·0–26·9 % from 2 to 4 d, by 6·5–16·4 % from 4 to 7 d and by 4·0–11·6 % from 7 to 12 d for energy and fifteen nutrients. In conclusion, the present study suggests that the day of the week, recording sequence and seasonality minimally contribute to the variation in nutrient intake. To measure Korean usual dietary intake using open-ended dietary instruments, 3–4 d may be needed to achieve modest precision (r>0·7) and 6–7 d for high precision (r>0·8).