When You Can Compute Any Metric, Which One Should You Trust?
Reading private-markets performance in the ILPA-transparency era
As of June 2026.
For a decade, an LP largely took the GP's headline number on faith. That era is ending. The ILPA Performance Template (released by ILPA in 2025) — standardizing gross and net IRR, TVPI and MOIC, and, notably, a transaction-level cash-flow table that lets an LP recompute performance under its own assumptions (including IRR with and without subscription-line impact) — moves the LP from reading a reported metric to deriving the metrics it actually trusts. It applies to funds commencing operations on or after January 1, 2026, with first deliverables expected in Q1 2027.
That is a real shift, and it raises a question the old "trust the GP's IRR" world let LPs avoid: once you can compute anything, what should you compute — and which number is load-bearing for the decision in front of you? The academic literature long ago established that no single metric is sufficient: the hazards of IRR are well-documented (Phalippou, 2008), and the public-market-equivalent (Kaplan and Schoar, 2005) exists precisely because absolute multiples ignore timing and the public-market alternative. The practical question — the one the new transparency forces — is how to reason across them. Three worked examples make the stakes concrete.
1. IRR and DPI can tell opposite stories about identical cash
Consider two funds. Each takes $100 from the LP and returns $200 in total — identical money in, identical money out. The only difference is when the cash comes back.
| Money in | Money out | DPI | TVPI | IRR | |
|---|---|---|---|---|---|
| Fund A — returns $100 at year 1, $100 at year 2 | $100 | $200 | 2.0× | 2.0× | 61.6% |
| Fund B — returns $200 at year 6 | $100 | $200 | 2.0× | 2.0× | 12.2% |
Same DPI. Same TVPI. IRR of 62% versus 12% — a fivefold gap created purely by timing. IRR is a rate: it rewards speed, and it implicitly assumes the LP reinvests early distributions at that same high rate — an assumption almost no LP can actually meet. DPI is immune to both effects; it is simply cash out over cash in. Neither is wrong — they answer different questions. If the LP cares about how fast capital came back, A wins decisively. If the LP cares about total realized return, they are identical.
2. A "bad" DPI can be a perfectly healthy fund
The instinct to swing from IRR to DPI runs into the opposite trap. Consider a healthy buyout fund three years into its life, $100 committed:
| Called | Distributed | Unrealized NAV | DPI | RVPI | TVPI |
|---|---|---|---|---|---|
| $60 | $5 | $75 | 0.08× | 1.25× | 1.33× |
A DPI of 0.08× looks alarming in isolation — and judging this fund on DPI alone would condemn it. But it is three years old: value is building (TVPI 1.33×), it simply hasn't been realized yet. This is the J-curve, and it is why DPI — for all its honesty about cash — is backward-looking and incomplete for any fund not yet near harvest. The right read pairs DPI (what's realized) with TVPI/RVPI (what's building), and weights them by where the fund is in its life.
3. The number you're now allowed to strip out: the subscription line
This is the distortion the ILPA template specifically targets. Take one investment that exits in year 4 at a 2.0× — DPI is 2.0× no matter what. But a subscription-line facility lets the GP delay calling the LP's capital by a year:
| LP capital funded | DPI | IRR | |
|---|---|---|---|
| Without sub line | Year 0 | 2.0× | 18.9% |
| With sub line | Year 1 | 2.0× | 26.0% |
The same deal reports a 7-percentage-point higher IRR simply because the LP's money was deployed for less time — with no change whatsoever in cash returned. This is not fraud; it is arithmetic. But it is exactly why the new template breaks out performance with and without the impact of fund-level subscription facilities — and why an LP that only ever saw the headline IRR was, unknowingly, comparing funds on inconsistent footing.
4. The blindness you cannot strip out: when IRR and MOIC agree and still mislead
The first three are distortions you can identify and correct for. The last is different — it is a structural blindness, and it survives even when every standard metric agrees.
Two LPs each decide that $100 of private-equity exposure is right for their portfolio. They use the same fund, with the same return pattern. But private capital is called gradually while early distributions flow back, so a commitment of $100 never puts a full $100 to work — a portion sits idle. Call that gap slack. One LP commits exactly $100 and lives with the slack; the other sizes the commitment up to actually deploy the intended $100.
| Commitment | Capital deployed | IRR | MOIC | Net profit ($) | |
|---|---|---|---|---|---|
| LP-Under (commits to the headline number) | $100 | ~$100 | 25.1% | 1.52× | $52 |
| LP-Right (sizes to hit the target exposure) | $152 | ~$152 | 25.1% | 1.52× | $79 |
Identical IRR. Identical MOIC. 52% more absolute profit. By every standard performance metric the two LPs are indistinguishable — yet one earned $27 more on the same intended allocation. This is not a timing trick or a financing artifact you can back out: IRR is a rate and MOIC is a ratio, and both are scale-invariant by construction — they measure return per dollar deployed, and so are structurally incapable of seeing whether an LP actually deployed the capital it set out to. The metric that would catch it — absolute dollars of realized profit against intended exposure — is not one the standard toolkit reports at all; it has to be computed deliberately, in dollars, against the allocation the LP actually set out to make.
So what should an LP compute?
The lesson across all four is not "use DPI instead of IRR," and it is certainly not a new single number to crown. It is that each metric is load-bearing for a different question — and some questions no standard metric answers at all:
- DPI — have we actually been paid? The cleanest realized-cash measure; the right anchor when liquidity is the concern. Incomplete for young funds.
- IRR — how fast? Useful, but timing-distorted and sub-line-sensitive; read it ex-subscription-line, and never compare two funds' IRRs without checking their cash-flow timing.
- TVPI / RVPI — what's the total, including what's unrealized? Necessary for in-life funds, but rests partly on the GP's marks — trust it in proportion to valuation discipline.
- PME — did the LP beat simply owning the public market, equivalently timed? The risk-and-timing-aware comparison the others lack; the closest thing to a "should we have bothered" test.
- Absolute realized profit vs. intended exposure — did we actually put the capital we meant to invest to work, and what did it earn in dollars? The question Exhibit 4 exposes — and the one no standard rate/ratio metric reports.
The discipline the new transparency rewards is not picking a favorite metric — it is reading them as a system, knowing which one answers the question actually being asked, and being explicit about what each one cannot see. The LPs who benefit from ILPA's raw cash flows will be the ones who compute deliberately, not the ones who simply swap one headline number for another.
This note reflects the state of play as of June 2026. The ILPA Performance Template (released 2025) applies to funds commencing operations on or after January 1, 2026, with first deliverables expected in Q1 2027. The worked examples are illustrative and arithmetically exact. References: Phalippou, L. (2008), "The Hazards of Using IRR to Measure Performance," Journal of Performance Measurement 12(4); Kaplan, S. N., and A. Schoar (2005), "Private Equity Performance," Journal of Finance 60(4).