Seconds, hours, days — and what a “month” has to mean

30.436875 days — exactly one twelfth of a Gregorian year of 365.2425 days, which works out at 2,629,746 seconds. It is an average, not a month you can point to on a calendar: no real month is 30.436875 days long. It is used because the alternative is choosing 30 or 31 arbitrarily, and a flat 30 days is right for four months of the year and wrong for the other eight. An honest average beats a number that is silently wrong two thirds of the time.

Because the Earth does not take a whole number of days to go round the Sun. The Gregorian calendar absorbs the leftover with a leap day every four years, then skips it in century years, then puts it back every fourth century — 2000 had a 29 February, 1900 did not, and 2100 will not. Averaged across that 400-year cycle every year is 365.2425 days, which is the figure used here and the reason a year comes to 31,556,952 seconds rather than 31,536,000.

86,400 — that is 60 × 60 × 24, and it is exact. A week is 604,800. Neither needs a convention, because for the purpose of measuring a duration a day is defined as that fixed count of seconds. Daylight saving and leap seconds change what a clock reads on a particular date; they do not change what “one day” means as a length of time.

It comes out as 2.9569 months, because the average month here is 30.436875 days rather than a flat 30. If what you actually need is the date three months from today, that is a calendar question and no fixed number of days answers it: three months after 30 November is 28 February, which is 90 days in a common year and 91 in a leap year. Durations and calendar spans are different questions and this page only answers the first.

They are ignored here, as they are in every general-purpose converter. A leap second is occasionally inserted into UTC to keep atomic clocks in step with the Earth’s slightly irregular rotation. Counting them would change a year by at most one second — far below the uncertainty already introduced by averaging months — and would require knowing exactly which years you meant.

Because they are nothing but the year multiplied by ten and a hundred, and they would carry the same averaging caveat scaled up with them. Multiplying the year figure yourself keeps it obvious which convention produced the answer, which is worth more than saving one keystroke.

No. The arithmetic runs in your own browser, there is no server behind this site, and this page writes nothing to your device.