How the Monitor is built

The Monitor tracks institutions (banks, central banks, CSDs and ICSDs, exchanges, custodians, asset managers, payment and stablecoin issuers, and the technology providers behind them), the specific initiatives they run, and the jurisdictions they operate in. Each institution carries a current tokenization status, a scored profile, dated developments (“signals”), and the sources behind every material claim.

Coverage is deliberately global and anti-concentration: the roster is not allowed to collapse onto the usual US / UK / CH / SG / HK / EU cluster.

Every material value is a claim, and every claim is graded independently on two axes (after the NATO Admiralty Code): the reliability of the source and the credibility of the information. Sources sit in a fifteen-tier hierarchy — legislation, central-bank and regulatory filings and audited statements at the top; press releases and executive statements in the middle; vendor case studies, conference decks and social media at the bottom.

A claim supported only by low-tier, marketing-origin sources is capped as not verified. Each institution also carries a 0–100 evidence-quality score — a function of how much of its profile is confirmed by high-tier sources, corroborated across independent reporting, and recent. That score is both a scoring dimension and the driver of how wide its confidence interval is.

Each institution is scored 0–100 on ten dimensions against anchored rubrics (anchors at 0 / 25 / 50 / 75 / 100). The weights are fixed and published:

The composite is a weighted geometric mean of the ten dimensions, not an average. Additive scoring is fully compensatory — a strong ecosystem story can paper over weak regulatory or commercial reality. Geometric aggregation is non-compensatory: a genuinely weak dimension drags the whole score down, and cannot be bought back by strength elsewhere.

No composite is ever presented as a bare number. Each carries a confidence interval — derived by perturbing the weights and resampling within the evidence-derived bounds — and a graded confidence level of high, moderate or low. Thin or low-tier evidence widens the interval.

Alongside the 0–100 composite, each institution gets a maturity level from 0 to 10 — from no identified activity, through research, consortium participation, pilots, and limited production, up to a token-aware operating model integrated into core infrastructure.

Crucially, overall maturity is capped by demonstrated commercial and operational reality: one high-profile pilot cannot, by construction, produce a high overall level. Failed and discontinued initiatives are kept permanently — the record of what did not work is part of the intelligence.

The hardest part of this field is telling adoption from announcement. These gates do that:

• Non-production events (an MoU, a sandbox admission, a hackathon, a single demo) can never, on their own, justify a production status.
• A single transaction is a test — “production” requires repeated activity on distinct dates with distinct counterparties, or a regulator / FMI confirmation of go-live.
• Consortium membership alone caps an institution at “consortium participation”; a 40-member network is a weak signal, not adoption.
• Represented value (operational recordkeeping) is never aggregated with distributed value (wallet-transferable) — the two are reported separately.
• Announced capacity is never mixed with realized activity.
• Marketing-tier sources can never be the sole evidence for a live, production or scaled claim.
• Republication date is not activity date — events are dated from the underlying event.
• A status upgrade requires new evidence dated after the previous status.

Every institution records when it was last verified, its latest known activity, and a freshness class. The Monitor is refreshed by feeding the tracked roster to a web-capable research process, which returns fresh signals, updated facts and re-scored profiles; each batch is previewed and applied under a locked, versioned methodology, and every change is audited. Scores always reference the methodology version they were produced under.