Working days per month mistakes: the traps and the way around them

Monthly counting errors survive review because they produce plausible numbers. Nothing raises an alarm at 21 when the answer is 23. The error shows up two quarters later as a rate that will not reconcile.

Each item below is a distinct failure with a distinct symptom. They are not variations on one point.

What to take away

Using an average month as a divisor

A year of weekdays divided by twelve is about 21.75, and that figure gets rounded to 22 or to 21 and then used as though it were a month. It is neither. No month has 21.75 weekdays, and the months that have 23 are not rare.

The symptom is a daily rate that reconciles at the year and never at the month. If a monthly total is being spread per working day, the divisor has to be that month's own count.

If the same rate must hold all year, say so explicitly and accept unequal monthly totals. Do not force them equal with an average that describes no month.

Not knowing which day the month starts on

Two 31-day months in the same year can differ by two weekdays. That is the entire mechanism, and it is invisible if you look only at the length. March and October in 2027 are both 31 days; one has 23 weekdays and the other 21, because one starts on a Monday and the other on a Friday.

The check takes a second: a 31-day month starting Monday, Tuesday or Wednesday has 23. Starting Thursday it has 22. Starting Friday or Saturday it has 21, and starting Sunday, 22. The derivation is under the monthly formula, and it is short enough to hold in your head.

Adding one to make it inclusive

Spreadsheet functions that take two dates already include both ends. Adding one to the result of such a function does not fix an inclusive-exclusive mismatch, it creates one.

The reliable test is a span you can count on your fingers. A Friday to the following Monday is two working days if both ends count, and one if only the end counts.

Whichever your tool returns is its convention. Write everything downstream to that convention instead of correcting it later.

This is the off-by-one problem in its usual disguise: the error is one, so the result stays plausible.

Treating February as always 20

A 28-day February has 20 weekdays for every possible start day, which makes it the one month where the shortcut is safe. That is exactly why it is unsafe: the habit survives into a leap year, where a 29-day February has 21 weekdays unless it starts at a weekend.

The leap rule itself is the ordinary one, and the century exception is the part people forget rather than the four-year rule. It is set out under the leap year rule, and a monthly model that hard-codes February at 28 will be wrong once every four years in a way that no test written in an odd year will catch.

Counting a closure that fell at a weekend

A closure on a Saturday removes nothing from a weekday count, because the Saturday was never in the count. Subtracting it anyway takes a real working day off the month.

This is common because closure lists are usually published as dates, and a list of dates does not announce which of them fall at a weekend. If your calendar applies a substitution rule, the substituted date belongs in the list and the original does not. If it does not substitute, the weekend date simply has no effect.

Storing the closure list inside the formula

A monthly count with dates typed into the formula cannot be re-run for a different calendar, cannot be audited, and cannot be updated without editing every month. Hold the list in one range with a publisher and a retrieval date beside it, as set out under the closure-list discipline, and let every month read the same range.

Mixing days with hours

A month that lost a half day to an early closure did not lose a working day. If a downstream unit is hours, the day count is the wrong carrier for that information.

Hour arithmetic handles that under working hours. Converting between the two by multiplying by eight is a separate assumption. Write it down where someone can disagree with it.

Comparing two months that used different calendars

Two offices in one organization can count the same month differently and both be right. Comparison only makes sense once the calendar is stated.

A monthly series that switched calendars mid-year has a step that looks like a business change but is not one. The same problem at the whole-year level, where it is larger and easier to see, is under counting a year.

Common questions

Is there a safe default monthly figure?

No. If a single number is genuinely required, use the year's total divided by twelve and label it as an annual rate, not as a month.

How do I check a monthly count I was given?

Ask for the month's start day, its length, and the closure list with its source. Those three reproduce the answer. Without the third, the count cannot be checked at all.

Do these errors matter at the scale of one month?

A two-day error in a 22-day month is nine percent. Whether that matters depends on what the number feeds, but it is not a rounding difference.

Which of these is the most common?

The average divisor, by a wide margin, because it is the only one that feels like a simplification rather than a mistake.