How the Numbers Are Made

Ravariant produces one verified number that other software can rely on: the loss that lenders actually took on a group of similar private funds, signed off by auditors. We publish two things. The first is a conforming facility standard, the checklist a loan has to pass. The second is PARI, the Private Asset Risk Index, the published loss figure, one per cohort, where a cohort is a group of similar funds, for example PARI.PC.SR for US private credit senior. A single public fund document goes in at one end and one PARI loss number comes out the other, and what follows are the 6 steps in between. The number is a fact about what already happened rather than a guess about what comes next, and the steps are arranged so that other software can rely on it without trusting anyone at Ravariant to have judged it.

Two design choices run through every step. The fund and its manager hand us nothing, so nothing in the figure depends on a party with an interest in how it reads. And the figure is a recorded fact produced automatically, so no one at Ravariant decides its value or its timing.

Classify into a cohort

Production starts with a single document, the fund's public offering memorandum, which we read once and use to sort the fund into a cohort. Funds land in the same cohort when they lose money for the same reasons, which is the whole point of the exercise, because losses are only worth comparing across funds that fail in the same way. There is no call with the manager, no questionnaire, and no consent sought or needed. The cohort, not the individual fund, is the unit the loss index is computed over.

Sorting carries this much weight because funds in one cohort run on the same engine. A trade receivables fund earns when businesses pay their invoices and suffers when they do not, while a buyout fund lives and dies on selling companies and on the cost of borrowing. The sales language differs sharply from one fund to the next and the underlying engine barely does at all, which is why the losses recorded across a cohort line up the way they do. That cohort is the address a conforming loan, defined in Step 2, reports its loss into.

Define the conforming facility

The index measures the loss recorded on loans made against fund interests, so before it can measure anything it has to fix which loans count. The conforming facility is the checklist a loan must pass to belong, and a loan that fails any single item is left out. A blended loss figure is only readable when every loan inside it has the same shape, since otherwise the reader cannot tell whether the number moved because losses rose or because the mix of loans changed. The checklist is public and identical for every loan in a cohort.

Loan to value is the load bearing rule of the set, fixing how much is lent against each dollar of fund value. A conforming loan sits in the lowest layer by design, so the index ends up measuring loss on first in line loans that all share one known shape. The standard is the first of the two things Ravariant publishes, and the band a loan stays inside is market practice, set per strategy and positioned by the realized loss read, which the next page works through.

The advance rate band by strategy

PARI does not set the advance rate. The advance rate is what NAV lenders actually do, a market practice band per strategy, driven by illiquidity and soft marks rather than by the realized loss PARI measures. What PARI publishes is the realized loss, and that loss read positions a cohort inside its band: a cleaner loss history sits toward the high end of the band, a stressed reading sits toward the low end. Different private credit strategies carry different risk and so trade around different bands.

The bands sit inside what the NAV lending market actually advances. The Financial Stability Board notes that loans against fund interests generally lend near 30 cents on the dollar today, and across the market the advance runs roughly 10 to 20 cents on each dollar of fund value, near 14 on average, with asset based lending reaching higher.

Senior secured direct lending is the best supported strategy, because it has the realized loss history that positions where inside its band a cohort sits today. That reading is elevated right now, so senior credit sits near 16 inside its 15 to 25 band. The other strategies show market practice bands as references, and each gains a positioned read once that strategy has its own verified loss history. In every case PARI publishes the loss and the band stays market practice; PARI does not derive the advance rate.

Verify the input

The index is only as trustworthy as the loss figures fed into it, which is why a lender's word on its own does not clear the bar. A figure an outside auditor has signed off on cannot be talked up or talked down by the party reporting it, and that is the whole reason the audit sits in the path. A committed lender feeds its audited realized loss on the NAV facility it holds, a Big Four or equivalent auditor confirms both the loss and that the loan exists, and the fund submits nothing at all. What we publish is the lender's loss blended across the lenders in a cohort, with the manager nowhere in the loop.

What counts as a loss

Only a realized credit loss counts. The fund genuinely cannot repay the facility, the collateral is enforced and worked out, recovery comes in below the loan, and the audited shortfall is the loss. We exclude three things because each one is gameable: a voluntary sale of the loan in the secondary market at a discount, a soft revaluation writedown, and anything that does not breach the conforming line, the loss level where coverage starts. What remains is a real credit event and nothing else.

How much a cohort can carry depends on what its lenders feed. Coverage runs in three tiers, a coverage tier being how verified the data is, level 0, 1, or 2, and only the top tier is settlement grade.

Where the underlying fund also trades through a public vehicle, its public price gives a second read on the audited figure, but that read serves as a guardrail rather than the source. The audited loss is what makes the published index a fact rather than a guess, since a loss an outside auditor has signed off on is simply what happened, and what happened does not later turn out wrong. Every step that follows is built on it.

Compute the index

PARI blends the audited losses across a cohort into one figure weighted by size, so each conforming loan contributes its audited loss and a larger loan counts for more than a smaller one. A single fund's loss would expose that fund and swing around too much to settle anything against, whereas the blended figure for the group holds steady enough for two parties to rely on it. The output is the settlement number, one recorded loss figure per cohort, assembled entirely from facts already signed off in Step 3.

Breadth rules sit on top of the weighted average. A cohort publishes only once it holds at least 4 to 6 members, and a single name cap limits how much any one loan can drive the figure. Together these keep the number credible and broad enough for the rules we operate under, and they make it hard to attribute a move to any one fund.

The single name cap and the minimum member count together let the index publish without exposing any one lender or fund. A reader sees the cohort loss, not the loan that carried it, which keeps Ravariant a publisher of a shared fact rather than a commentator on any single fund.

Calibrate the standard

Calibration reads the realized loss from history, and it is where whatever modeling we do lives. The loss read comes from the strategy loss history and the cohort of peer managers, not from any single fund's books, because where a cohort sits inside its band has to be grounded in how badly the strategy has actually behaved rather than picked by feel. The advance rate band itself stays market practice; calibration positions the cohort inside it. This is an internal reading and not a published prediction.

A loss engine driven by one main factor sets the dials. It is fed the realized loss history of comparable public lenders, plus an allowance for rare large shocks so the tail is not undersold. From that the engine reads an everyday expected loss and a worst case loss in the gravest 1 in 200 years.

The advance rate band we track for each strategy is the NAV facility advance rate, the share of fund value a NAV lender advances, a market practice band rather than a number PARI derives. The realized loss read positions a cohort inside that band, and for senior secured direct lending the reading is elevated today, which puts the line near 16 cents on the dollar inside its 15 to 25 band. That line is what we measure the lenders' losses against.

These limits are set once and revised only through the governed version process in Step 6, and calibration itself outputs no live signal, ranking, or flag. The only numbers Ravariant publishes remain the standard and the recorded loss index built against it.

Version and govern

A number that contracts settle against has to be trustworthy in a way anyone can check, since a settlement figure is only worth using if a party who disputes it can reproduce it and arrive at the same answer. US bank regulators wrote down what that takes in guidance known as SR 11-7, which holds that such a number be documented, tested against history, and reviewed by someone who did not build it. Ravariant holds the conforming standard and the loss index to that same discipline.

Every number carries its version and its date

Each published value records which version of the method produced it and the date it speaks for. When the standard or the index method changes, the version number ticks up and a change log records what changed and why. Numbers published under earlier versions stand as published.

A rulebook resolves disputed numbers

When a published index value is disputed, a written rulebook governs the outcome rather than a judgment call. Source data is stored exactly as received, so any cohort value can be rebuilt for any past date using only what was known at the time. Disputes are settled by rebuilding the number against the rulebook, not by anyone's memory.

Because the index is a recorded fact and the standard is a published rule, governance here means keeping the record permanent and the number reproducible rather than defending a guess. There is no prediction that could come out right or wrong, only a number that can be replayed and a rulebook that says how.

PARI runs itself

None of the 6 steps is run by hand each period. PARI is a published ruleset that runs on its own, and given verified inputs the number is a pure function of them, the same every time and reproducible for any past date. Nobody at Ravariant chooses the value, the timing, or when a series rolls; the rules choose all three, which is what keeps the publisher impartial.

The series and the roll

PARIX, the tradable instrument built on the PARI number, settles on what these loans actually lose. A cohort trades as a dated one year series, for example PARIX.PC.SR.2026, fully funded up front and cash settled. Each series carries its loans to the end, settles once on the loss they recorded, and expires. A holder rolls into the next series to stay on, and no position runs without a settlement at the end.

Two speeds, the interim mark and the audited truth

The number runs on two clocks at once. The administrator strikes a NAV monthly, which gives a live interim mark labeled as such, while the annual audit strikes the realized loss, the hard figure the series settles on, and the interim marks square back to it once a year. PARI moves monthly and settles on fact annually.

When a loan fails

When a conforming loan takes a real audited loss, the lender that took it, not the fund, files the audited shortfall, and that loss enters its cohort and the index rises. The lender files fast, because its own coverage pays out against the number it reports. The market moves even sooner, as the monthly NAV marks fall and the tokens reprice ahead of the audit. A soft writedown or a voluntary sale at a discount does not move it.

Publication parameters

The six steps produce the standard and the index; these are the rules they publish under. Each is a fixed, named parameter a party can rely on and a contract can reference.

The pattern is the one index publishers settled on decades ago. The method is public, the parameters are named, changes are announced in advance, and the published record is permanent. That is what lets a party wire the index into a contract once and rely on it without rereading this document every quarter.

What the index does not publish

Ravariant publishes two things, the conforming facility standard and the recorded audited loss index. Five other kinds of output are left out on purpose, and each omission is part of what keeps the published number a fact systems can settle on.

No prediction of trouble ahead

Ravariant does not publish a guess about where stress is heading. A guess can turn out wrong; a loss an auditor has signed off on already happened.

No ranking or flag

There is no score from 0 to 1 for where a cohort sits and no NORMAL, WATCH, or STRESS label. The index reports the recorded loss itself, built from audited facts.

No price for fund interests

Nobody can price a private fund interest from public data, and Ravariant does not claim to. The index measures recorded loss on conforming loans, not the worth of any fund interest.

No letter grades or pass and fail verdicts

The index does not squeeze a cohort into a symbol or declare a fund sound or unsound. It publishes the recorded loss and the conforming standard, so each party reaches its own verdict within bounds its own committees approve.

No buy or sell opinions

An opinion would make the publisher a player in the trade. A recorded number with no view serves both sides at once, which is what lets two parties settle against the same value.

The document ends where the thesis begins. What we put out is not a guess about a fund's risk; it is a published standard and a recorded audited loss, signed off by outside parties and built to be settled on.