Rules
Part of Which parts of May birthdays deserve the closest look?
When is extra scrutiny worth it for May birthdays?
May birthdays by day: the number stored under every date, what a spreadsheet does to a day column, and the checks that survive an import pipeline intact.
What to take away
- A May birthday is only as trustworthy as the day, month and year that survived every hop from the keyboard to the page.
- Reopen May 1, May 31, records where both numeric fields are 12 or lower, records with a two-digit year, and anything that arrived as text.
- May has no leap-day test, so a shifted month or a shifted century will not announce itself.
- Verify a disputed date against a birth record, not against another birthday list.
Which May records need the closest look
Most wrong May birthdays were typed correctly. The damage comes later, when a program reads a correct value under a different convention and rewrites it. Six signals justify a recheck.
- May 1 and May 31.Each edge sits one conversion away from the wrong month. A May 1 stamp saved at midnight UTC shows as April 30 at 8 p.m. in US Eastern time.
- Both numeric fields 12 or lower.05/06/1990 is May 6 under month-first convention and June 5 under day-first.
- The month arrived as the word May.May is also a first name and a surname, so a names column or a notes column can be converted without anyone asking.
- A two-digit year between 30 and 49.Excel reads 35 as 1935; Microsoft SQL Server reads the same 35 as 2035.
- A row with no leap-day check.February 29 fails loudly outside a leap year. May has no matching test.
- A month column sorted as text.Alphabetically May falls between March and November, not April and June.
A month page can only claim as much as the month field proves, which is what the May birthdays index works through before it lists a name.
A sample May calendar, by day
The table below is a test fixture, not a finished index. Years appear as they are commonly recorded.
| Date | Recorded entry | Year as recorded | What to check |
|---|---|---|---|
| May 1 | Wes Anderson | 1969 | Month edge; a midnight UTC stamp shows as April 30 in US Eastern |
| May 4 | Audrey Hepburn | 1929 | Baseline row, nothing structural to fail |
| May 6 | George Clooney | 1961 | 5/6 also reads as June 5 in day-first locales |
| May 13 | Stevie Wonder | 1950 | A day above 12 fixes the field order |
| May 24 | Bob Dylan; Queen Victoria | 1941; 1819 | One day, two centuries, so the year decides everything |
| May 29 | John F. Kennedy | 1917 | 5/29/17 returns 2017 under a 00 to 29 window |
| May 31 | Clint Eastwood | 1930 | June 1 at midnight UTC shows as May 31 in US Eastern |
If an import turns the Kennedy row into 2017, the pipeline is guessing and every row is suspect. A record that carries a day and a month but no year is a separate case.
Example: one typed day, three results
Autoconversion is easiest to watch in a column of days pasted from a notes file into a spreadsheet.
What Spreadsheets Do to Days
Typed
- Day and month, no year
- date this year
- Two-digit year
- wrong century
- Day above twelve
- text or date
- Leading zero
- zero stripped
- Name like date
- converted to date
Comes back
- Day and month, no year
- Two-digit year
- Day above twelve
- Leading zero
- Name like date
One typed day, three results
What can come back
- A day and month with no year
- A date in the current year
- 5/6 with no year
- May 6 or June 5
- A year typed as 35
- 1935 in Excel, 2035 in SQL Server
- A name resembling a date fragment
- A date
Why
- A day and month with no year
- The parser fills the missing field instead of refusing
- 5/6 with no year
- Field order is a locale setting
- A year typed as 35
- Windowing rules differ by product
- A name resembling a date fragment
- Autoconversion never reads the column heading
Format the destination column as text before you paste. Formatting it afterwards does not help, because the conversion is already done and the original characters are gone. Partial dates, where the year was never typed, need a second decision, and the January birthdays by day page works through those cases.
The number under the date
Software does not store a date as the words on the page. Excel counts days from January 1, 1900 and calls that day serial 1. Google Sheets and LibreOffice Calc count from December 30, 1899, one day earlier. Unix counts seconds from January 1, 1970 at 00:00 UTC, the epoch.
Move a column between two systems without declaring the format and the whole column shifts by a constant. One worked case is enough: May 1, 2024 is Excel serial 45413 and Unix 1714521600.
The only way to catch a constant shift is to carry rows whose correct value you already know. A May 1 entry and a May 31 entry will fail loudly when a column moves.
Two-digit years and the century guess
The year 2000 problem was fixed in the systems people worried about, not in every export or hand-kept sheet. Excel maps 00 through 29 to 2000 through 2029, and 30 through 99 to 1930 through 1999. Microsoft SQL Server splits at 50 instead.
Two-Digit Year Century Shift
Written
- Year field
- two digits
- Day and month
- valid May day
- Record age
- intended
- Rendering
- looks fine
Read as
- Year field
- guessed century
- Day and month
- valid May day
- Record age
- off by 100 years
- Rendering
- looks fine
A stored two-digit year is not a compressed date. It is a date with a missing field and a guess standing in for it. February is the one month where a wrong day fails on its own, and the February birthdays by day write-up shows what that failure looks like.
A verification pass in five steps
- Set the destination column to text, and declare the import format rather than letting a parser infer it.
- Load known-good May rows, including one May 1 and one May 31 entry, whose correct values you already hold.
- Export what you imported, compare it with the source file, and inspect the rows that differ.
- For a row that still disagrees, order the record. The CDC page on where to write for vital records explains how to request a birth certificate by state.
- Log the time zone and the field order beside every May import, so a day 1 or day 31 record can be rechecked without rebuilding the pipeline.
Once a row is confirmed wrong, rebuilding it is a separate job, and the January birthdays update shows what a verified month set should look like.
