We have developed a method of estimating day of ovulation using urinary ovarian hormone data. The method identifies a day of luteal transition that occurs at the shift from production of follicular oestrogen to luteal progesterone. The algorithm for identifying this shift was evaluated and judged better than specified alternatives in that it resulted in (1) a high concordance between the day of luteal transition and peaks in urinary luteinizing hormone (LH) for cycles with well-defined peaks, (2) a low variance in the length of the luteal phase of the menstrual cycle, which presumably reflects a low measurement error in estimating day of ovulation, and (3) a high proportion of cycles for which an approximate day of ovulation could be determined. To validate the new algorithm, it was applied to an independent data set. The algorithm identified a day of luteal transition in 88 per cent of these cycles, and the identified day occurred within two days of the urinary LH peak for all the cycles with clear LH peaks. Determination of the day of luteal transition to estimate ovulation requires only first-morning urine specimens, requires no correction for day-to-day variations in urine concentration, and can be applied to a mid-cycle window of data.