This question, posed in a recent comment to my Fooled by IRRs post, deserves an answer in the form of a post. It has made me realize that the inaugural post of my blog, The Quartiles’ Oxymoron, was not as self-explanatory as I thought it was.
Quartiles are meaningless and benchmarking IRRs against an IRR benchmark is not an apples for apples comparison.
The academic version of the answer (Damodaran et al.) is that IRRs should not be used to rank investments because of its flaws (multiple solutions, etc.), therefore not even IRR vs IRR is an apples for apples comparison. For subjective (because of the discretionarily chosen discount rate) benchmarking, NPV is the way to go.
From my practitioner’s standpoint, cash in- and out-flows have an impact on duration and on the notional capital amounts being compared in a way that is not accounted by IRRs, as the different simplified cases of the table below, all yielding an identical IRR, show.
In other words, an 11,34% IRR does not unambiguously mean that at the end of observed period the initial capital is grown by 11,34% compounded annually (which is true only in the case of Deal A of the table above), and actually it does not in case of interim distributions (when it means that the invested capital has grown 11,34% for a duration shorter than the observed period).
In addition, early distributions, whilst they usually enhance IRRs (i.e. they require a lower multiple to achieve the same IRR – as in Deal D and E of the table), yield cash returns rather than (supposedly higher) ROIs (causing reinvestment risk).
GPs would usually opine that freed cash can be directed to other investments – true in theory, but concretely possible and beneficial only if the next investment is a good one and if it is available for an immediate lump-sum cash injection (that is never the case). And yet, all this is about a different investment than the one that is being benchmarked.
Eventual reinvestment opportunities (implicit in the IRR calculation) should not bias performance measurement.