How we arrive at a number

The source hierarchy

Every figure on this site is labelled with where it came from. We use three tiers, and we never silently mix them.

1. Regulatory filings. SEC EDGAR — Form N-MFP3 for registered money market funds, Form D for private placements, prospectuses and statements of additional information. The GLEIF legal-entity register for domicile and corporate structure. These are the strongest evidence available and they win against everything else.
2. The issuer’s own disclosure. Product pages, factsheets, documentation, published APIs. Strong, but marketing pages go stale and sometimes contradict the same firm’s filings — we have found exactly that more than once.
3. Third-party trackers. rwa.xyz, DefiLlama, CoinGecko. Useful for coverage and for cross-checks, weaker as evidence. Where a row rests on a tracker we mark it tracker in the table rather than letting it look issuer-verified.

What every figure must carry

A number without a date is not a measurement, it is a rumour. Our data model makes this structural rather than aspirational: every value is stored together with the date it refers to, the source name, the source URL and its evidence tier. A field we cannot source stays empty and renders as n/a. There is no code path that produces an estimate.

Note the distinction between the date a figure refers to and the date we happened to fetch it. We record the former. Where a source publishes a live value with no stamp at all — several issuer dashboards do this — we say so instead of inventing one.

Yields are not one thing

This is the single most common error in tokenized-treasury comparisons, including in tables published by firms that should know better.

A 7-day SEC yield is a defined output of Rule 2a-7 and exists only for US-registered money market funds. On our board, that is BENJI/FOBXX and WTGXX — both file Form N-MFP3 monthly and report a real one. BUIDL is a British Virgin Islands 3(c)(7) fund, USYC a Cayman fund, USDY a debt note rather than a fund at all, and USTBL a French UCITS under EU money-market rules. None of them files under 2a-7, so none of them has a 7-day SEC yield to publish. What they publish instead is a computed APY, a 30-day yield, or nothing.

Putting all of those in one column headed “yield” compares measurements that are not the same measurement. We therefore carry the metric name beside every figure on the Yield Board, and we would rather show an empty cell than a number that looks comparable and is not.

When we publish nothing

• When the issuer does not disclose it. BlackRock and Securitize publish no yield for BUIDL. A tracker shows 1.47% and 0.34%, which are implausible for a Treasury fund and appear to be an artifact of the fund paying dividends monthly as newly minted tokens — a mechanic that a naive price-change calculation cannot read. We show no yield for BUIDL.
• When a chart would be decoration. We have no sourced twelve-month series for the market as a whole, so we do not draw one. An invented trend line is a fabricated measurement, whatever the axis says.

When sources disagree, we publish the disagreement

A reader who knows that two trackers differ by 31% on the same day is better served than one shown a single confident number. Every conflict we find is listed on the Yield Board with both claims, both links, and how we resolved it — or a note that it is still open. Some of what we found:

• Circle’s own product page states assets of roughly $313M for USYC, about a tenth of the figure returned by Circle’s own live API on the same day.
• Franklin Templeton’s website API reports a 7-day yield of 4.57% while the firm’s own N-MFP3 filings have been flat between 3.43% and 3.60% every month from December 2025 to June 2026.
• A widely used tracker describes OUSG as domiciled in the British Virgin Islands. Its SEC filing and Ondo’s own compliance page both say Delaware limited partnership.
• One tracker’s single “Spiko” entry of $2.23B is 6.7% the US Treasury fund. The rest is a euro T-bill fund, a sterling one, and a UCITS whose return comes from total return swaps over an equity basket.

What counts as a tokenized treasury

We include a product only where the holder has a claim on a portfolio of short-dated US government paper. We exclude, and say why we exclude:

• Stablecoins that hold tokenized treasuries. Counting Ethena USDtb beside BUIDL double-counts the same underlying bills, because around 90% of its reserves are BUIDL.
• Unsecured corporate debt. A face-amount certificate is an obligation of the issuing company backed by that company’s assets, not a pro-rata claim on a segregated portfolio.
• Non-US government paper bundled into the same tracker entry.

Where a product is genuinely borderline — a fund of tokenized funds, or a long-duration ETF wrapper sitting in a table of cash-equivalents — we keep it and mark it, rather than dropping it silently.

Review cadence, and how we handle being wrong

The board carries the date of its last full end-to-end re-check, not the date the page was rebuilt. Every row is re-sourced at review, not carried forward on trust.

We will get things wrong. When we do, we correct the page and say what changed rather than editing quietly. One example already: an earlier draft of this site asserted that none of these products files under Rule 2a-7 and that no 7-day SEC yield therefore exists for any of them. That was wrong — BENJI/FOBXX and WTGXX are registered 2a-7 government money market funds and do report one. It was caught before publication by checking the filings rather than trusting the generalisation.

How we link out, and why we do not rehost

Every citation on this site opens the original document, in a new tab. We do not download third-party studies, filings or factsheets and serve them from this domain, even where the licence would allow it. Two reasons, and the second is the important one.

The first is copyright. Some of what we cite is genuinely reusable — EU documents are available under CC BY 4.0 by Decision 2011/833/EU — but most is not. A rating agency’s fee disclosure, a vendor’s price list and a central bank’s bulletin each carry different terms, and “it was useful to us” is not one of them.

The second is that a copy goes stale. If the European Central Bank corrects a figure in its bulletin, our copy becomes a frozen wrong version — on a site whose entire method is that a number carries the date it was read. Hosting someone else’s document would recreate exactly the failure we spend these pages documenting in others.

Outbound links are also not a leak to be plugged. On a reference page, sending a reader to the filing is the product. What we do control is whether a link passes a ranking signal: citations to regulators, registers, courts and published research are followed, because nofollowing the SEC on a site whose claim is “we read the filings” would be an incoherent signal. Links to the commercial parties we compare — issuers, platforms, vendors — carrynofollow, because a neutral comparison should not hand ranking signals to the companies inside it.

Independence

We take no payment from any issuer, platform or protocol for inclusion, for placement, or for coverage. There are no affiliate links on this site. Full terms are on the disclosure page.