The IRR is calculated by a trial and error process
Starting with a guess at the IRR, r, the process is as follows:
- The NPV is calculated using discount rate r.
- If the NPV is close to zero then r is the IRR.
- If the NPV is positive r is increased.
- If the NPV is negative r is decreased.
- Go back to step 1.
This is more tedious than calculating an NPV, but the extra work can be automated. It avoids the need to estimate an appropriate discount rate which is a considerable simplification — this is a flaw because of the lack of risk adjustment. This can be done later by demanding risk premia for higher risk alternatives, but assessing this reintroduces the complexity.
It is generally preferable to use NPV to IRR to make investment decisions. A smaller investment with a better rate of return will have a higher IRR, but investors' total wealth would be increased more by making a larger investment with a lower IRR but a higher total gain.
This is only a problem where investments are limited in size (not scalable) and mutually exclusive, so it is not a concern for securities valuation. However, the only common use of IRR for securities valuation is that of yield to maturity for bonds.
IRR has even worse failings. If the investment has negative cash flows following positive cash flows, then there may be more than one IRR, or even none at all. While it is possible to use more complex procedures to work around this, it is better to simply not use IRR. If used, it should not be used for any pattern of cash flows that ever changes from positive to negative.
Some of the failings of IRR are addressed by MIRR.