Start with the signed ledger, timing basis and NPV at the hurdle. IRR summarizes a zero of that valuation function; it cannot alone rank projects of different scale, duration or risk.

Periodic timing versus calendar dates

Periodic IRR is effective per selected interval. Dated XIRR uses actual elapsed days divided by 365 and is annual effective, even across a leap year. Do not compare a monthly rate directly with an annual hurdle. Time zero belongs in NPV; the worked examples below retain annual year-end timing.

Investment, financing and mixed flows

A conventional investment pays out first and receives later; financing receives first and pays later. Under the conventional pattern, their IRR comparisons run in opposite directions. Mixed sign changes can produce several roots. Read the NPV sign at the chosen hurdle, with approximate break-even near the scaled tolerance; it is not a universal recommendation.

Finite search and numerical sensitivity

The rate domain stays above −100%. Read the displayed search bounds and each candidate’s residual. A finite search is not proof that all roots exist within it or were found. No root found, invalid data and nonconvergence are different states. Near-flat profiles can make a reported rate sensitive even when its residual is small.

MIRR and the optional illustration

MIRR makes financing and reinvestment assumptions explicit. Its dated version uses elapsed Actual/365 years and is a separate extension, not Excel MIRR. Neither MIRR nor a wealth illustration proves that funds can earn the assumed return. A project IRR must not become an automatic reinvestment forecast.

Comparable projects and private records

Use matching currencies, valuation dates, hurdle assumptions and included costs. MOIC ignores timing; PI uses discounted inflows and outflows. Keep sensitive account identifiers out of the ledger. Check shared figures and exported raw flows, dates, rate basis, root bounds, residuals and MIRR assumptions before distributing a report.