Days between: two answers, both right, and the one that loses you a deadline

Work it out in DATES, which gives both figures side by side and names them, rather than picking one and hoping. It also counts working days against your own holiday list, and says when it clamped a month rather than rolling it. Nothing is uploaded.

The two counts

Take Monday the 21st and Wednesday the 23rd.

Three conventions, all standard, differing by one day at each end. Whichever you use, somebody else is using a different one, and the difference is invisible in the answer: 2, 3 and 1 all look like perfectly reasonable answers to “how many days?”

Where each one is used

Days elapsed is plain subtraction, and it is what you want for anything that accrues: interest, age, time served, how long something has been outstanding. It is also what a computer gives you if you subtract two dates and do not think about it, which is why it is the answer people accidentally use everywhere.

Days involved, sometimes called inclusive counting, is what you want when each day is itself a unit of the thing being counted: nights in a hotel, days of cover, days of leave, days a vehicle was off the road. If you are charging or paying per day, this is nearly always the count.

Clear days is the one people have usually never heard of and is used throughout court rules and statutory notice periods. “Not less than seven clear days' notice” means seven days that are neither the day of service nor the day of the hearing. Serve on the 1st for a hearing on the 9th and you have given seven clear days; serve on the 1st for the 8th and you have given six, and the application fails on a technicality that has nothing to do with its merits.

Why this loses deadlines

Because the error is exactly one day, and one day is the size that survives a sanity check. If a miscount produced an answer of forty days when the truth was three, somebody would notice. An answer of seven when the truth is six looks completely fine, gets written into a diary, and is discovered at the hearing.

It is also the kind of error that compounds quietly across an organisation: one person counting inclusively and another elapsing, both confident, both showing their working, disagreeing by a day and each assuming the other has made an arithmetic slip.

The defence is not to memorise which convention applies where. It is to write down which one you used, next to the answer. A diary entry that says “14 days from service, counting the day of service, expires 3 October” can be checked by somebody else. One that says “3 October” cannot.

Adding months, which has no single right answer at all

What is one month after the 31st of January?

There is no fact of the matter. The 31st of February does not exist, and the two obvious repairs disagree:

• Clamp: take the last day of the target month, so 28 February (or 29 in a leap year).
• Roll: carry the overflow forward, so 3 March.

Both are defensible and both are implemented in real software. Spreadsheets typically clamp. Some scheduling systems roll. Contract terms often specify, and where they do not, the clamping answer is the more common reading of “one month later” in English usage — a month after the last day of January is the last day of February, not a few days into March.

The practical consequence is that “three months from the 31st” is a phrase to avoid in anything that matters. Write the date. If you cannot, say which convention the parties intend.

Note also the asymmetry: clamping is not reversible. One month after 31 January is 28 February; one month after 28 February is 28 March, not 31 January. Adding and then subtracting a month does not return you to where you started, which breaks any process that assumes it does.

Working days, and why nobody can ship you a holiday list

Counting working days needs three things: which days of the week are working days, which specific dates are not, and whether the count is inclusive.

The first is not universal. Monday to Friday is common and far from global; a Friday–Saturday weekend is standard across much of the Middle East, and plenty of operations run a different pattern internally.

The second cannot be supplied by software honestly. Public holidays differ by country, differ by region within a country — England and Scotland do not share a calendar — differ by year, and differ again by whether your organisation actually closes. A tool that shipped a built-in list would be quietly wrong for most of the people using it, and quietly is the problem: nothing on the screen would say which list it used.

So the list has to be yours, pasted in, and the count has to say how many days it excluded and why.

Week numbers, which are not what people think

The ISO week numbering defines week 1 as the week containing the first Thursday of the year, with weeks running Monday to Sunday. Equivalently: the week containing the 4th of January.

The consequence surprises somebody every January. If the 1st of January falls on a Friday, Saturday or Sunday, it belongs to the last week of the previous year — so the 1st of January 2027 is in week 53 of 2026. That is not a bug, and a system that reports “week 1” for it is using a different and less standard definition. If two systems disagree about the week number of a date in early January, this is almost always why.

The habit worth having

Write the convention next to the answer. Every time. It costs four words and it is the only thing that makes a date calculation checkable by the person who has to rely on it.