Quick answer
Work through the list: birthday passed, endpoint counting, month-length borrowing, end-of-month clamping, decimal divisor, time zone. One of them explains nearly every discrepancy.
Off by one year
The single most common discrepancy. One tool subtracted birth year from reference year; the other checked whether the birthday had actually occurred. For anyone whose birthday falls after the reference date — about half the population at any moment — those give different answers.
Diagnostic: compare the month and day of birth with the reference date. If the birthday is still ahead, the higher figure is the wrong one.
Use this guide with the right tool: Open the Age Calculator for exact chronological age in completed years, calendar months and remaining days. If the question shifts, compare it with the Date Difference Calculator or the Inclusive Date Count Calculator.

Off by one day
Inclusive versus exclusive endpoint counting. Elapsed counting measures midnights crossed; inclusive counting counts both endpoint dates. The difference is exactly one, permanently, and both are legitimate.
Time zones produce the same symptom for a different reason: a tool using local time and one using UTC can disagree about which calendar date it currently is, which shifts the reference date by a day for several hours out of every 24.
Off by one to three days in the days field
This is month-length borrowing. When the days column goes negative, a correct method borrows the actual length of the preceding month; a common shortcut borrows thirty. February borrows give the largest divergence.
End-of-month clamping produces a related symptom. From 31 January, one month forward is 28 February under clamping and 3 March under rolling. Any sequence of monthly steps built on different conventions drifts apart from the first step.
Decimal answers that nearly match
Decimal ages depend on the divisor: 365, 365.25 or 365.2425. Over a few decades those give visibly different figures for the same interval, and none of them is wrong — they are different definitions of an average year.
If two decimal ages differ in the second or third decimal place, the divisor is almost certainly the cause. Check the stated method rather than recalculating.
What to do about it
Compare stated methods before assuming one tool is broken. A calculator that documents its convention is more trustworthy than one that produces a number with no explanation, even when the undocumented one happens to agree with you.
Where a threshold or an official decision depends on the answer, the convention that matters is the one specified by the governing rule, not the one your preferred calculator uses. Find the rule, then pick the tool that matches it.
Worked example
For 18 November 1994 on 16 August 2026, completed birthdays give 31, subtracting years gives 32, and elapsed versus inclusive day counts give 11,594 or 11,595 days. Run your own dates in two tools and compare their conventions before deciding which answer fits your purpose.
One person, five defensible answers
The worked example puts two disagreements on the table for someone born 18 November 1994, measured on 16 August 2026. The full set of readings for that one pair of dates shows how many answers can be defended at once. Completed years: 31. Year subtraction, 2026 minus 1994: 32. Running age — the person is in their 32nd year of life: 32 again, agreeing with the wrong method for a different reason. The calendar structure: 31 years, 8 months and 29 days. Total days: 11,594 elapsed, or 11,595 counted inclusively. And the decimal readings of the same 11,594 days: 31.7644 dividing by 365, 31.7426 dividing by 365.25, and 31.7433 dividing by 365.2425. The 365 divisor disagrees in the first decimal place; the other two part company only in the third. Five questions were being asked, and each answer is the right response to one of them.
The days-field disagreement deserves a full working, because it is the one that looks most like an error. Take a birth on 25 January 2026 and a measurement on 10 March 2026 — 44 elapsed days. Worked forward: one complete month reaches 25 February, and thirteen days remain, giving 0 years, 1 month, 13 days. Worked by columns with a flat thirty-day borrow: the days column is 10 minus 25, borrow thirty, giving a days field of 15 and an answer of 0 years, 1 month, 15 days — two days out. Borrow February's real length, 28 days, and the column method lands on 13, agreeing with the forward count. The tools are not disagreeing about the dates; one of them borrowed a month that does not exist.
Clamping produces the widest structural split of all. From 31 January 2026 to 30 March 2026 is 58 elapsed days. With end-of-month clamping applied step by step, one month reaches 28 February, a second reaches 28 March, and two days remain: 0 years, 2 months, 2 days. With rolling steps, one month from 31 January lands on 3 March, and the remainder is 27 days: 0 years, 1 month, 27 days. The totals agree to the day; the structures share almost nothing. If two tools return these two answers for a January 31st birth, neither has miscalculated — they have made different, documentable choices about what a month means at the end of January, and the stated method is the only place the difference can be settled.
The calendar reading can be checked against its anchors: 18 November 1994 plus 31 years is 18 November 2025, plus 8 months is 18 July 2026, plus 29 days is exactly 16 August 2026 — the structure reassembles the reference date, which no year-subtraction answer can do. And the conventions stack. A tool that counts inclusively and divides by 365 reports 31.7671 for this same person on this same day: the day-count convention and the divisor convention, each defensible alone, compound into a figure that matches no other tool on the page. When a disagreement reaches the third significant figure, stop comparing answers and start listing conventions — there will be one per layer of the difference.
Common mistakes to avoid
- Assuming the tool that agrees with you is correct. Check the stated methods. A documented convention beats a familiar answer.
- Recalculating instead of comparing methods. The arithmetic is rarely the problem. The convention almost always is.
- Picking a convention to match a desired outcome. Where a rule governs, its convention applies regardless of what any calculator does.
Use the right calculator
Age Calculator gives exact age in years, months and days, Date Difference Calculator gives calendar difference and total days, and Inclusive Date Count Calculator counts days with start and end inclusion stated.
Frequently asked questions
Which conventions does this site use?
Completed years, elapsed day counting, real month-length borrowing, end-of-month clamping, and 365.2425 for decimals. Each page states what applies.
Is one convention more correct?
Not in general. Correctness comes from matching whatever rule governs the decision the number feeds.
Why do phone and web calculators differ?
Often time zone handling. A device using local midnight and a tool using UTC disagree about today's date for part of each day.


