Methodology: how we classify, mark and measure
Last updated: September 5, 2026
TokenizedXonChain runs The Cube — a neutral newsroom and data-intelligence desk for tokenized equities and real-world assets. We are not a broker-dealer, exchange, issuer, custodian, or investment adviser; we do not execute trades, hold customer funds, or mint or redeem tokens. This page states exactly how every number we publish is computed, which source it came from, how often it refreshes, and what we do when a source fails. Everything here is informational only and is not financial advice.
Last reviewed: September 5, 2026. This document is re-reviewed whenever a formula, threshold, source or refresh cadence changes. The constants quoted below are read directly from the modules the site runs on, so they cannot drift out of date independently of the code.
The registry · claim classes
What a token is: the claim class
Every wrapper carries one class. It says what the holder can actually enforce, against whom. Only an issuer-native token is the share; every other class references it.
| Class | What the holder has | Same as the stock? |
|---|---|---|
| Issuer-native share | The token is the security, recorded on the issuer's own ledger. Holding it is holding the share. | yes |
| Custodial entitlement | A claim on shares a custodian holds for you. The shares exist; your name is on the custodian's books, not the company's. | no |
| Tracker certificate | You are a creditor of the vehicle that holds or references the share. You are not the shareholder of record. | no |
| Linked note | A debt instrument whose payoff follows a share. You are owed money by the issuer, not a slice of the company. | no |
| Synthetic | A payoff contract that moves with a price. Nothing is held for you. | no |
| Deposit token | A bank deposit represented on a ledger, used to settle. | no |
| Fund share token | A share in a fund (money-market, T-bill) recorded on a chain. BUIDL and Benji are this. | no |
| Unclassified | The desk has not signed this row's class yet. A ticker suffix is not a class. | not signed yet |
Fan tokens and club equity tokens are classes too, but they live on the SportFi book and never appear in the tokenized-stock registry.
The registry · who decides
Signed by the desk, never inferred
A class is set only by a desk signature, per issuer family or per row. A ticker suffix (on, x, c) is a hint shown to the desk, never a class. A row the desk has not signed says so; it is a visible hole, not a guess.
- Ondo Global Markets: linked note. The holder is a creditor of the issuer, not a shareholder.
- xStocks (Backed Assets, Jersey): tracker certificate on the share.
- Coinbase tokenized stocks: custodial entitlement to a share held in custody.
- Securitize (SECZ): issuer-native share, recorded on-chain by the issuer.
Family positions signed 28 Sep 2026. Every change to a class is written to an audit trail with the time and the signer, and fires a class-change flag in the Office.
Marks · the number we would defend
The mark
The mark is the NAV while it is fresh. When it is stale, the mark falls back to the on-chain last and is labelled on-chain, NAV stale. With no NAV at all it is labelled no official NAV. The on-chain print is never silently called NAV.
Marks · the stale policy
When a NAV is stale
- US cash session open: a NAV older than 30 minutes is stale.
- Cash session closed: a NAV is stale if it is not from the last session (the close at 20:00 UTC).
Stale values stay on screen, marked; they are never blanked or zeroed.
The registry · rights
Vote, dividends, redemption
For each wrapper the desk records whether the holder votes, how dividends reach them, whether and when the token can be redeemed, and whether the issuer of the share objected. Fields the desk has not signed read “not signed yet”. Since the 17 September 2026 Innovation Exemption these rights are a pricing input, not a footnote.
The registry · collateral
Where a token is taken as collateral
Posting tokenized stocks as collateral is the use case that is actually shipping. We record each venue with its status (live, announced, paused) and flag listings and removals. Collateral is not ownership: what can be enforced at 03:00 on a Sunday depends on the class.
Microstructure · measured, not assumed
Does the token discover the next open?
We test it on every name: an ordinary least-squares fit of the next cash open's move on the token's return while the cash market was shut. The best public study (Dombrowski and Arshadi, 2026) finds a slope near zero; we publish our own estimate with its sample size, never a borrowed conclusion.
The Cube · news policy
Desk and Wire
Every story names its source. Desk stories are written by hand at this desk. Wire stories are drafted automatically from a cited feed and labelled so. Each story also carries one label for what it changes for a holder: wrapper, register, collateral, pipe, regulator, SportFi, deposits or data vendor.
SportFi · a separate book
Fan tokens are not equity
Fan tokens are utility tokens; the announced club equity token is a security with no offer and no price yet. The two are never summed, and neither is mixed into the tokenized-stock registry or its totals. See the SportFi desk.
Live status · read at page render
What the price tape is running on right now
Pyth Hermes, our low-latency price oracle, became key-only on 26 August 2026. Keyless requests return HTTP 401. Until a Pyth key is configured in this deployment, equity quotes on TxOnChain are not Pyth prints and are not real-time exchange quotes. We state that plainly rather than implying a live oracle.
Prices: fallback (no Pyth key). No Pyth key is answering, so nothing on the tape is a Pyth print. Equity rows are licensed Ondo on-chain token prices — current, and labelled by their own age, but prints of the token rather than real-time exchange quotes for the underlying share — or a delayed public cash quote where Ondo did not answer. Crypto rows are CoinGecko, upgraded to the on-chain Chainlink round when that round is under 65 minutes old.
One precision about that third chip: pricesAsOfis the newest Pyth publish time when Pyth answered at all. When no Pyth entry is present, there is no oracle publish time to report, so the field carries the moment the response was assembled — which is why per-symbol freshness is judged from each print’s own publish time on the tape, not from this single aggregate stamp.
The source chain, in order
- Pyth Hermes with a key — the licensed low-latency feed. Labelled LIVE.
- Pyth legacy keyless path — retried in case public access is restored. Labelled LIVE if it answers.
- Pyth crypto-only retry — a single non-entitled feed ID rejects the whole batch, so when a key is present but partial we re-ask for the crypto subset alone. That is what produces the
mixedstate: genuinely live crypto, non-Pyth equities. - Ondo Global Markets — for equities, the licensed on-chain print of the tokenized instrument. These rows are the token (NVDAon, AAPLon), not the NYSE cash quote, and the tape names them that way.
- CoinGecko, upgraded by Chainlink— for crypto. We overwrite the CoinGecko price with the on-chain Chainlink round when that round is less than 65 minutes old — the aggregator’s one-hour heartbeat plus a small tolerance, because a round that has not moved enough to re-publish is still the current on-chain price. Older than that, the round is genuinely stale and we keep CoinGecko rather than serve an old on-chain answer.
- Delayed public cash quote — last resort, only for equities none of the above answered, stamped with the source market time so it is labelled DELAYED.
Nothing in that chain is a mock. If every step fails, the affected row shows a dash. We do not substitute a cached or invented figure.
One thing this chain does not do is turn a tokenized print into an exchange quote. When the tape reads LIVE beside NVDAon while Pyth is unavailable, it is telling you the token’s on-chain price is current — not that you are seeing a real-time NYSE quote for NVDA. Those are different instruments, and the premium formula in the next section exists precisely because they can and do diverge.
Implementation: src/lib/publicApis/pyth.ts · src/lib/publicApis/fallbackPrices.ts · flag published at /api/stats.
Formula · multi-oracle consensus
Multi-oracle consensus and divergence
One price feed is a claim. Several independent feeds agreeing is evidence. The consensus rule is deliberately simple, so it can be checked by hand:
divergence_i = |price_i − avg| ÷ avg
consensus = strong when max(divergence_i) ≤ 0.3% · alert above
A source that did not answer is excluded from the mean rather than counted as zero — a missing feed must never drag the average down. Severity on an alert is tiered: above 0.5% is medium, above 0.8% is high.
Which sources can be present — and which are today
| Source | What it is | Status today |
|---|---|---|
| Ondo GM | Licensed on-chain token price and issuer NAV for the tokenized equity | Present — the primary equity source |
| Finnhub · Tiingo IEX | Two independently licensed cash quotes for the underlying share, each with its own timestamp | Present for the most traded names; the independent side of the consensus |
| Pyth | Independent oracle price for the underlying, scaled by the shares-multiplier | Key-only since 26 August 2026. While our key is refused the row reads key rejected and contributes nothing; the delayed public quote that stands in on the ticker is labelled as delayed, never as Pyth |
| Chainlink | Keyless on-chain reads of Ethereum mainnet: the feeds DeFi uses for Ondo tokens (an “Ondo API” relay and Chainlink's own “Calculated” price), the 24/5 stock feeds, and the crypto majors | Presentbeside Ondo's print on the Ondo page and on each asset page. The 24/5 stock feeds count as an independent mark in the Office consensus; the Ondo relay feeds never do |
| Hyperliquid XYZ | Equity perpetual futures on the same underlying | Present as another market pricing the stock; a derivative, never an input |
| RWA.xyz | Issuer-attested reference data | Dark until a key is supplied — contributes nothing |
Chainlink's Ondo token feeds and Hyperliquid perps never enter a NAV, premium or consensus figure. Chainlink's Ondo feeds are shown beside Ondo's own print because they are what a lending market reads when it values the token as collateral, and one of the two relays Ondo's price, so it cannot be an independent check of Ondo. A perp pays the price move and nothing else; it has no NAV.
So, plainly:the consensus in the Office checks the issuer against people who are not the issuer. The issuer's own per-share mark (Ondo underlying × multiplier) is compared with the median of the independent cash quotes, and the issuer is never averaged into the number that checks it:
gap_i = (mark_i − ondoMark) ÷ ondoMark × 10,000 bps
dispute when any |gap_i| > 100 bps · awaiting when no independent mark answered
A row marked awaiting is honest about the limit: it has one source, so it is a single claim, not a consensus. While the stock market is shut the cash quotes are the last close, so the gap is the overnight move, and it is labelled that way.
Implementation: src/lib/oracle.ts (canonical rule) · src/office/desk/panels/DeskMultiOracle.tsx (the Office table) · src/lib/ondo/chainlinkFeeds.ts and src/app/api/chainlink/feeds/route.ts (on-chain reads) · src/lib/perps/hyperliquid.ts (perps).
Rule · divergence and de-peg
When a gap becomes a de-peg
A premium or discount is normal. A de-peg is a large, persistent break from NAV that signals liquidity, oracle, or backing stress. We are careful about the difference, and careful about what we can actually prove from the data we hold.
The risk score
The NAV board scores each asset 0–100 from three observable components, and shows the score beside the row:
momentum part = min(25, |24h change%| × 5)
liquidity penalty= 20 when reported on-chain token volume < $5,000,000, else 0
risk = min(100, sum of the three) · amber ≥ 30 · red ≥ 60
The liquidity penalty is there because a wide premium on a thinly-traded token is a different thing from a wide premium on a heavily-traded one — the first may simply be nobody trading, the second is the market disagreeing with the backing. The volume input is the on-chain token volume reported by the venue tickers feed, which covers the venues in that feed and not the whole market.
What we do not claim
These are point-in-time readings on the latest observation, not a confirmed sustained divergence. The public surfaces do not keep a continuous per-asset price history, so we cannot — and do not — assert that a gap has persisted for a fixed number of minutes. We therefore never label an asset de-pegged from a single reading. A gap is worth attention when it is wide, when it shows up across independent sources rather than in one feed, and when it is still there on the next refresh; that judgement is made by a person, not by a timer.
Everything here is informational.Scores, buckets and divergence flags are observations about data quality and market structure. They are not signals, recommendations, price targets, or advice, and they never trigger an action on anyone’s behalf — TxOnChain has no execution path of any kind.
Implementation: src/components/panels/NavPremiumHeatmap.tsx (risk score) · src/lib/oracle.ts (buckets and severity tiers).
Operating rules · cadence and failover
Freshness, cadence and failover
A number can never be fresher than the cadence of the source behind it. We match our caching to each upstream’s own update rhythm rather than pretending to a speed the source does not have.
| Source | Cadence | Notes |
|---|---|---|
| Ondo GM prices | 5 s | Office panel tick. Shared server cache of 5 s, so the platform quota stays flat however many readers are on the site. Server-rendered pages such as /data re-read the NAV board at most once a minute, and say so on the board |
| Ondo GM — own key | 1.5 s | A reader-supplied key bypasses the shared cache and runs on their own quota |
| Ondo market / status | 1 min | Heavier payload; server cache 1 min |
| Ondo metadata | 5 min | Contracts, chains, corporate-action reference data — near-static |
| Chiliz transfer tape (Ankr) | 6 s | ERC-20 Transfer events on the registry’s frozen CC2 fan-token contracts, read from Chiliz Chain over Ankr HTTPS by a dedicated ingest worker. The site reads the worker’s snapshot; pages never call the RPC. Amounts are token units — no USD notional is estimated. Marks and TVL stay on the sources above; Ankr never feeds either |
| Pyth Hermes | 10 s | Short memo, deliberately; data older than the memo is never re-served, it is re-fetched or dropped |
| Non-Pyth quotes | 60 s | CoinGecko, Chainlink override and delayed cash quotes share one 60-second memo |
| Chainlink on-chain | 30 s | Aggregator read cached 30 s; the server-rendered data desk re-reads at most every 5 min |
| DefiLlama | hourly | TVL, stablecoin supply and yields move slowly; an hourly read is the honest cadence |
| Rates curve | daily | Treasury par yields and the effective fed funds rate publish once per business day; cached one hour |
| Site aggregates | 60 s | /api/stats — asset count, RWA TVL, stablecoin supply, tape prices |
The DELAYED rule
Every price carries its own publish time from the source, and we label from that time, not from when we fetched it:
- LIVEThe print’s own publish time is less than 10 minutes old. A continuously-ticking feed — the Ondo token prints, a Pyth update, a fresh Chainlink round — is stamped at read time and so passes this test. LIVE is a statement about the age of that specific print and nothing more.
- DELAYEDThe publish time is 10 minutes old or more. The newsroom tape carries one label for the whole strip, so it is marked DELAYED unless at least half of its symbols are fresh by that same 10-minute test.
- DashNo source answered for that symbol. We print a dash, not a last-known number.
What happens when a feed dies
- We never re-serve a stale print as liveThis is not theoretical. When Pyth went key-only in August 2026, a cache kept replaying the same 26 August prints for eight days while the interface still said LIVE. The fetch layer was rebuilt so that a failed upstream returns nothing at all instead of the last successful response, and the caller then either falls through to a labelled fallback or renders an empty state.
- Failed reads are never cachedAn empty or errored result is not written to any memo, so the next request retries the upstream rather than inheriting the failure for the length of a cache window.
- Sections disappear rather than lieA server-rendered board whose source returned nothing is omitted or replaced with an explicit unavailable note. No placeholder rows, no invented figures, no last-year’s-number standing in for today’s.
- Age is always availableWhere a value is shown with a source, its as-of time is shown with it, so a reader can judge staleness themselves instead of trusting our word for it.
Cadence figures on this table are imported from src/lib/ondo/config.ts and the route handlers themselves, so they change when the code changes.
Standards · independence and production
Editorial independence
The Cube covers the tokenized-asset sector neutrally. We hold a read-only data licence from Ondo Global Markets, which is not an endorsement. We accept no paid placement inside editorial or data coverage, no affiliate links in editorial coverage, and we hold no position in the assets we report on. Our job is to describe what is happening accurately — not to promote any issuer, token, or venue.
How the newsroom is produced
The Cube uses an automated editorial desk that monitors public reporting and primary sources — issuer announcements, regulatory filings, and established crypto and financial media — then drafts an original, long-form article on each genuinely new, in-scope story. Every article names the source it paraphrases, never reproduces source paragraphs verbatim, and uses at most one short attributed quote per source. We deduplicate by underlying event, so the same story is not republished under reworded headlines, and we do not pad, fabricate, or invent figures, dates, or quotes. When there is nothing genuinely new and in scope, we publish nothing.
Sector Tape · taxonomy v2.0.0
How a token gets a sector
A sector board dies if tickers hop categories week to week, so the taxonomy is versioned and published. Every tile on the tape is produced by the rules below — this page renders them straight out of the classifier, so what you read here is what ran.
The sectors
| Code | Sector | What qualifies | In RWA ex-stables |
|---|---|---|---|
| S1 | Tokenized equities | Single-name stocks, ETFs, equity baskets and 1:1 or synthetic wrappers delivering equity exposure. | yes |
| S2 | Tokenized funds | Money-market, bond, multi-asset, private-equity and crypto fund shares. The wrapper is a fund, not a single stock and not a T-bill-only vehicle. | yes |
| S3 | Tokenized credit | Private credit, structured credit and tokenized loans. Kept separate from funds: different risk, redemption and buyer. | yes |
| S4 | Tokenized commodities | Gold, silver and other metals, plus commodity ETPs where the claim is commodity exposure. | yes |
| S5 | Tokenized treasuries & bills | T-bill and short government-debt exposure: direct bills and T-bill-only vehicles, as the source classes them. Money-market funds sit in S2. | yes |
| S6 | SportFi | Fan tokens, club equity tokens and Chiliz-native sports assets. Utility and equity claims are never blended. | yes |
| S7 | Other RWA | Real estate, carbon, collectibles and other real-world claims that fit no sector above. | yes |
| L0 | Settlement · stablecoins | Fiat-referenced tokens used as cash on-chain. Shown as rails, never as the headline number. | no · rail |
Classification rules
| Rule | Statement | Enforced by |
|---|---|---|
| R1 | Stablecoins never sit inside asset-sector tiles. They are the settlement layer and are excluded from RWA ex-stables. | classify() + checkInvariants() |
| R2 | An ETF wrapper of equities is EQUITIES, not FUNDS — unless the issuer is a money-market or treasury fund product. | classify() Ondo branch |
| R3 | A SPY-like index ETF token is EQUITIES. | classify() Ondo branch |
| R4 | A GLD-like token is COMMODITIES when the mandate is metal; EQUITIES when it is simply another equity ETF. | classify() Ondo branch |
| R5 | Robinhood Stock Tokens are EQUITIES even where they are legally debt notes. Legal form is a field, not the sector. | wrapper_type field |
| R6 | Money-market fund shares (BUIDL, BENJI, USYC) are FUNDS. A vehicle the source classes as T-Bills (OUSG, USTB, TBILL) is TREASURIES & BILLS. | classifyLlamaRwa() |
| R7 | Private-credit pools, CLO funds and reinsurance pools are CREDIT, kept separate from funds: different risk, redemption and buyer. | classifyLlamaRwa() |
| R8 | Socios club tokens are SPORTFI / fan_utility. The Socios Equity Token is SPORTFI / equity_token. The two are never summed. | classify() sportfi branch |
| R9 | Two wrappers of the same underlying ticker are SEPARATE assets that roll up to one underlying_id. | underlyingId() |
| R10 | The same contract dual-listed on two chains is one asset with multi-chain balances. | assetId() |
| R11 | Wrappers are not counted: a vault, rebasing or savings token holds an asset already counted, or a stablecoin. Bridged copies are one asset with balances on several chains. | classifyLlamaRwa() + buildTape() chain mix |
| R12 | Where sources disagree on size by more than 5% or $1M, the figure is flagged DISPUTED and both numbers are shown. | buildTape() disputes[] |
| R13 | Governance tokens, platform tokens and assets the source flags as suspicious are not real-world assets and count nowhere. | classifyLlamaRwa() |
| R14 | Size is on-chain market cap: circulating supply on each tracked chain times the source's reference price. Assets are placed by the source's own category and asset-class fields. | buildTape() + classifyLlamaRwa() |
Where the numbers come from
Sector sizecomes from exactly one source: DefiLlama’s RWA dashboard, which measures each tokenized asset’s on-chain market cap (circulating supply on every tracked chain times a reference price). The desk reads it hourly and places each asset by its own category and asset-class fields. One source is used not because it is the best figure for each sector, but because it is the only way the parts provably add up to the whole. A board that sourced equities from one vendor, funds from another and commodities from a third would print six numbers that reconcile to no headline at all.
Sector health — NAV premium, how many wrappers exist, how many actually traded — comes from Ondo Global Markets under a read-only data licence. That split matters: a generalist board can tell you a sector is big, and this one can tell you whether it is working.
One number we deliberately do not use: the market capitalisation Ondo reports per asset is the underlying company’smarket cap — Apple’s, not AAPLon’s. Summing it across the wrapper universe would state trillions of “tokenized equities”. The honest figure is the tokenized value itself, a few billion dollars.
Reconciliation
Every dollar in the dataset must land in exactly one place: a sector tile, a declared exclusion (wrappers, platform tokens, assets the source flags as suspicious), or a declared pile of assets the taxonomy could not classify. Stablecoins and governance tokens arrive only as totals and never reach a tile. The tape computes the residual on every build and publishes a reconciliation_residual gap if it is ever non-zero, and the tiles are checked to sum to the headline within $1 before anything is shown.
When sources disagree
Our RWA ex-stables total is lowerthan DefiLlama’s own RWA headline, and that is deliberate. DefiLlama counts wrappers (Sky’s sUSDS, Ethena’s sUSDe, rebasing and vault tokens) on top of the assets and stablecoins they hold; rule R11 does not. Rather than quietly pick whichever number is larger, the tape prints both, names the difference and explains it. Disagreements above 5% or $1M are flagged DISPUTED (R12) with both figures shown. Other trackers (RWA.xyz, Token Terminal) publish larger totals again because they count assets this source does not list; those are different measures and are never blended into ours.
History, and what we will not invent
This desk records one immutable close per UTC day. A day that has rolled over is never rewritten, and no history is ever backfilled — which means that on any given day the change columns only show horizons a recorded observation actually supports. A four-day-old series will not present itself as a 30-day move; the cell stays blank until the record is genuinely that old. Closes recorded under an earlier methodology stay at their dated pages but are never differenced against current ones, because a change of method is not a market move. Where a longer view is useful, DefiLlama’s own daily series is shown beside ours and labelled as theirs. Growth-quality grades read NA across every sector because grading growth requires net subscription and redemption flows, which this desk does not collect. The invariant check rejects any tile that carries a grade without flow data behind it.
What we do not claim
We are not a broker-dealer, issuer, custodian or exchange. We do not mint, redeem or hold customer assets. We publish the tape. Sector sizes are a third party’s on-chain market cap, which is a different measure from the “distributed value” figures other registries publish; the two are never blended into one number. Nothing here is investment advice.
Changelog
- v2.0.0 (2026-10-03) — Sector sizes now come from DefiLlama's RWA dashboard: each asset's on-chain market cap, classified by its own category and asset-class fields. Around 2026-09-22 DefiLlama moved real-world assets out of its protocol TVL adapters; the per-protocol figures 1.x read kept answering with their last value, so closes from about 2026-09-22 to 2026-10-02 repeat stale numbers. The daily series restarts on 2026-10-03.
- v1.0.0 (2026-09-06) — First published taxonomy: DefiLlama RWA-category protocol TVL, classified by protocol.
Commitments · the hard limits
What we will never do
- Never execute, route, or facilitate a tradeThere is no order path, no venue connection, and no execution code in this product. We are a read-only data and media platform.
- Never take custodyWe do not hold customer funds or assets, we never see a private key or seed phrase, and we do not mint, redeem, or settle tokens.
- Never give personalised investment adviceNothing here is a recommendation to buy or sell anything, and nothing is tailored to an individual’s circumstances.
- Never accept paid placementNo issuer, venue, or protocol can buy coverage, a ranking, a data slot, or a favourable description. Any commercial relationship that affects what we publish would be disclosed on the page it affects.
- Never present mock, cached, or invented data as liveIf a feed is silent we show an honest empty state or a labelled fallback. A number on this site is either real and attributed, or it is not there.
Questions we get
The short answers
How is the NAV premium calculated?
Premium = (token price − NAV) ÷ NAV, reported in basis points (× 10,000). NAV is the underlying market price multiplied by the token shares-multiplier, so a token that represents more than one share is compared against the value of all the shares it represents. A positive number means the token trades above the value of its backing; a negative number means below.
Are the prices on TxOnChain live?
Some are, and the label is applied per print by age, not by pipeline name. Pyth Hermes became key-only in August 2026, so without a Pyth key nothing on the tape is a Pyth print. In that state the tokenized-equity rows are licensed Ondo on-chain token prints, which are current and therefore carry a LIVE age — but LIVE there means the token's own on-chain price is current, not that we are showing you a real-time exchange quote for the underlying share. Rows filled from a delayed public cash quote carry that quote's own market time and read DELAYED. Any print whose own publish time is 10 minutes old or more is labelled DELAYED rather than LIVE, and a symbol nothing answered for shows a dash.
Which oracles does the consensus view actually use?
The issuer's own per-share mark (Ondo Global Markets) is checked against the median of the independent marks that answered: Finnhub and Tiingo IEX cash quotes, Chainlink's 24/5 stock feed where Chainlink runs one, and Pyth, or while Pyth refuses our key the delayed public quote, labelled as delayed. The issuer is never averaged into the number that checks it. Chainlink's feeds for Ondo tokens and Hyperliquid's equity perps are shown beside Ondo's print as other markets, and never enter a NAV, premium or consensus figure. RWA.xyz stays dark until a key is supplied.
What happens when a data feed fails?
We either show the last good value with its age and a DELAYED tag, or we show nothing. A failed read is never cached and a stale print is never re-served as live. Sections whose upstream returned nothing render an honest empty state instead of a number.
Does TxOnChain execute trades or take custody?
No. TxOnChain is a read-only media and data-intelligence platform. It does not execute trades, route orders, hold customer funds, custody assets, mint or redeem tokens, or give personalised investment advice, and it does not accept paid placement in editorial or data coverage.
Accuracy · corrections and limits
Accuracy and corrections
We aim for accuracy over speed. If you spot an error in an article or a data point, email hello@tokenizedxonchain.com and we will review and correct it. Material corrections are reflected on the affected page.
Known limits, stated plainly.Independent cross-checks exist for a minority of the universe, not all of it. Issuer data is only as good as the issuer feed. Delayed quotes are delayed. Volume-based liquidity penalties use the venue data available to us, which is not the whole market. None of this is an audit, and none of it is a substitute for the issuer’s own documentation.
See the live data overview for the boards these rules produce, the glossary for how each metric is defined, and the Disclaimer and Terms for the full legal position.