How to analyze any RWA in 30 minutes

The market has three sizes

Three correct, disagreeing readings of "the market" in 2026: distributed value excluding stablecoins about 38.32 bn USD (August 2026); represented value 377.74 or 435.93 bn USD, two RWA.xyz subpages, same day, 3 September 2026; CoinGecko-methodology market cap 19.32 bn USD (data to 31 March 2026). None of them is wrong. Each counts a different set.

RWA.xyz introduced the distributed and represented split on 21 November 2025. Distributed assets can leave the issuing platform and move between wallets. Represented assets cannot. The multiplier between the two categories is about ten in 2026, and most of the represented side sits on one institutional network. A headline about a tokenization market above 400 bn USD is, in the main, a closed network rather than a market on which an order can be placed. Chapter 1 sets the split; this chapter only needs the working rule that follows from it.

Reclassification looks identical to a capital crash in a time series. About 36 bn USD on 31 December 2025, via Stobox, against 14.1 bn USD on 1 January 2026, RWA.xyz via MetaMask, reflects two taxonomy changes, 21 November 2025 and 17 March 2026, not a 60% collapse. Represented value swung over 100 bn USD between readings weeks apart in summer 2026 and fell 8.55% in 30 days to August 2026. The working rule: represented-value moves of more than a few percent a month default to a dataset change, not a flow.

Same-month tracker disagreement without any reclassification is ordinary. Three July 2026 distributed readings sat 20% apart, 31 bn, 33.1 bn and 33.5 bn USD. Tokenized Treasuries in May 2026 had two readings 38% apart, above 11 bn against 15.20 bn USD, from the same underlying data. The direction of the market does not depend on the tracker. The levels do. From about 5 to 6 bn USD in early 2025 to about 38 bn USD in August 2026 is a six- to sevenfold move on any tracker.

A five-level source hierarchy

Level 1, the offering document. Prospectus, private placement memorandum, issuance terms, redemption rules. Binds the issuer. Contains what is nowhere else: who may use the primary market, and under what process the issuer may change terms or end the product.

Level 2, the regulator filing. Securitize's Form 424B3 states that NAV supplied to smart contracts comes from the fund administrator or the transfer agent, and that the platform's assets are not freely traded on a secondary market and have no continuous market price. The SEC's 11 December 2025 no-action letter for the Depository Trust Company sits here too.

Level 3, the issuer announcement. Often true and always selective, because the sender picks the metric.

Level 4, the dashboard. A dated measurement whose filter and methodology are usually hidden.

Level 5, the secondary portal. Repeats a dashboard read late, with no methodology.

Most of the most granular figures from summer 2026, exact amounts, holder counts, thirty-day changes, were available only through tertiary sources quoting RWA.xyz with a read date. That citation proves the number sat on a dashboard that day. It is not an independent confirmation. The operational rule: a number from a secondary portal needs a fresh read from the source and the reader's own date, and if that read cannot be done the number is citable with a caveat, not citable.

Level 4 is not self-consistent. Total RWA holders on 3 September 2026: 3,230,169 on /networks against 3,298,816 on /platforms, a 68,647 gap on the same day. If the metric were a set of unique addresses, the two sums would match. They do not, so it is an aggregate of separate lists, probably without full deduplication across assets. RWA.xyz's formal definition of "Asset Holders" was never established.

Five questions the offering document answers that no dashboard does

Read them in this order, because the first section decides the value of the rest.

Who has primary-market access, at what threshold, in what redemption window, and what exit exists for a holder who is not a primary client. BUIDL requires qualified-purchaser status and a 5 mn USD minimum under Rule 506(c) and Section 3(c)(7), with redemption in a T+0 or T+1 window. OUSG, USTB and USYC operate around 100,000 USD. BENJI starts at 20 USD. USDY is built as Regulation S outside the United States. For xStocks, redemption is open only to professionals through the issuer; a retail holder can only sell on the secondary market.

Under what process the issuer can change terms or end the product. That clause decided the largest wind-down in this class so far. On 12 May 2025 Anchorage Digital announced the acquisition of Mountain Protocol and the orderly wind-down of USDM, disabling mint the same day. About 10,820 holders, cap under 50 mn USD. Rewards ran to 11 June 2025, then the rate was set to 0%. On 22 August 2025 the primary market closed and remaining reserves went into a Uniswap pool against another stablecoin. The instrument that had been a claim for 1 USD against an issuer regulated by the Bermuda Monetary Authority became a token backed by USDC and honoured by an AMM. The underlying bills were fine. The product disappeared.

The legal structure and asset segregation: entity, jurisdiction, the document that constitutes the claim, and whether assets are legally separated from the issuer's estate. A BUIDL buyer does not buy Treasury bills. A PAXG buyer does not take title to a bar.

Who is custodian, administrator, transfer agent, auditor.

The full fee decomposition. Net yield depends on management fee, custody, platform, transfer agent and redemption fee, not on the underlying's rate.

An answer of "not disclosed" is recorded as a result. More than four such answers makes a product unmeasurable, whatever the quality of the marketing.

Reading contract permissions in the order that actually matters

The order is the reverse of intuition. First: who can upgrade the implementation. That answer invalidates every other one, because if the code can be swapped, every answer about functions is temporary. Then the inventory of functions that move someone else's balance. Then today's holders of those roles. Then proof of use from event history. Chapter 5 is the worked reading; this is the checklist.

BUIDL: a single externally owned account owns the proxy and holds the MASTER role, able to swap the entire implementation via setTarget in one transaction with no timelock and no multisig, read 4 September 2026. seize requires the TRANSFER_AGENT role, held that day by one EOA and one contract. Event scan from 1 March 2024 to 4 September 2026: zero Seize, OmnibusSeize, OmnibusBurn, Pause and Unpause events, 208 Burn events, more than 1,000 Issue events. Zero events is confirmed for Ethereum only.

OUSG: no seize, forcedTransfer or freeze exists in the code. BURNER_ROLE was never granted to anyone, confirmed by scanning every RoleGranted event and calling hasRole(). Admin power sits on a 4-of-7 Safe for token admin, a 3-of-5 Safe for the KYC registry, and both a 4-of-7 and a 1-of-9 Safe for pause. Real power lives in the removable identity registry, not a freeze function.

The same product on different chains has a different attack surface. On Avalanche, BUIDL has seize and omnibusSeize but not forcedTransfer, setTarget, freeze or mint, with an empty EIP-1967 admin slot and an EOA owner with no code. On Solana it uses the Token-2022 permanentDelegate extension with a plain system-account upgrade authority. On Aptos, ds_token::seize is a Move entry function. Supply split, 4 to 5 September 2026: Solana 977.9 mn, Avalanche 564.2 mn, Ethereum 211.9 mn, Aptos 161.3 mn. Ethereum is about 11% of the four-chain total. An Ethereum-only permission audit describes a minority of the product.

Tool limits are part of the result. Avalanche's public node caps eth_getLogs at 2,048 blocks, so a full scan of 41,884,144 blocks needs 20,451 requests. Solana and Aptos usage was left unevaluated for lack of indexer access. "Not evaluated" is the correct answer, not "zero".

One function call settles an oracle's freshness

Eight production NAV oracles, read 4 September 2026 at 11:08 UTC, gave USCC 225 days stale with no error flagged, and ACRED never written (roundId = 0, answer = 0). The rest ranged 0.1 to 21.7 hours old. Seven of eight had maxPriceDeviation confirmed at 0, disabled. USCC's call reverts, so its state is unknown, not zero. Chapter 6 is the measurement. The teardown only needs the method: call latestRoundData(), read updatedAt, and stop if the consumer has not checked freshness itself. Push feeds do not stream. They write when a threshold or a heartbeat is met, and they return the last value with a true old timestamp.

NAV is computed off-chain by a fund administrator and only carried by the oracle. Daily publication is not daily valuation if the underlying is priced quarterly and interpolated in between.

An attestation tests a narrow claim at one moment, usually that circulating tokens are covered by the fair value of reserves, and does not confirm reserve composition, encumbrances or internal controls. An audit examines a full period. USDC: monthly attestation (Deloitte) plus weekly disclosures. USDT: quarterly (BDO Italia). BUIDL: annual fund audit (PwC).

Four numeric traps that share one root cause

A missing unit.

A number with no date, after represented value moved more than 100 bn USD between summer readings.

Represented value counted as if tradable. JMWH, 861 mn USD, listed on no DEX and no CEX. JSOY_OIL, 300 mn USD, mint and burn only. In the commodity aggregate they look like 1.16 bn USD of market.

A cumulative figure reported as a period figure. BENJI's 211 mn USD-plus of peer-to-peer transfers is a five-year total, not annual, 8.6% of the fund over five years.

Counting positions instead of value. A newly listed stock token at 763 USD counts identically to a billion-dollar fund. Tokenized stocks are about 77 to 79% of total RWA holders. Holders measure state. Active addresses measure use. Chapter 10 scores those tests at segment level; the teardown computes them for one product.

Turnover-to-AUM, with the ceiling caveat, because transfer volume includes mint and redemption: Securitize about 1.4× a year; tokenized stocks about 51×; tokenized gold about 71×; PAXG about 31×. Active-address-to-holder: Securitize 61 active against 1,872 holders, 3.3%. DeFi-use: 7.4 bn USD of RWA deposited in DeFi in the second quarter of 2026, still under 10% of tokenized value; Aave Horizon 0.43% of the distributed market. Many of the largest assets were never designed to trade. Absence of turnover is a specification for a product that does not promise liquidity, and a failure for a product sold as liquid exposure to an illiquid asset. The marketing of the two is often identical.

The 13-question framework, with a time budget

What it tokenizes. What rights the token actually gives. Where the underlying is and who controls it. The legal structure. How mint and redemption work. Who the client is. Where the yield comes from. The business model. Why this chain. The risks. Whether the product makes sense. Whether it has product-market fit. Whether it is distributed or represented.

Time budgets below are the author's estimate from how long each check took in the source material, not a separately measured figure. They add to about 30 minutes if the offering document is to hand and the chain is one the reader can query. They add to more than 30 if the document is missing, which is itself a result.