01
Research principles
Only publicly observable controls and documentation are evaluated.
Unknown controls are never inferred; absence of evidence is not evidence of absence.
A control the review could not resolve is excluded from the score rather than counted as a failure, and is reported through evidence coverage instead.
Customer or commercial status has no effect on scoring, and using Decoda does not increase a score.
Evidence coverage is reported independently from the controls score, and decides whether a protocol is ranked.
Every protocol in an edition is reviewed against the same dataset version, so no asset is compared against a stale snapshot.
02
Observed Controls Score
The headline figure is the Observed Controls Score: what the controls the review could actually observe scored, not a verdict on the whole register. Each control has a published weight and an evidence-state multiplier. The score is the weighted points earned divided by the total applicable weight, multiplied by 100 and rounded to the nearest whole number.
Score=round(Σ ( weight × multiplier )Σ ( applicable weight )× 100)
Numerator — points earned: each scored control's weight times the multiplier of its evidence state.
Denominator — applicable weight: the weights of the scored controls. NOT APPLICABLE and UNVERIFIED controls are removed from both.
Score = round( Σ(weight × multiplier) / Σ(applicable weight) × 100 )
Controls marked NOT_APPLICABLE or UNVERIFIED are removed from both numerator and denominator. An unresolved control is not a failed control, so it is excluded rather than scored as zero; NOT_OBSERVED is a finding and does score zero. Where every control in a scope is unresolved there is no denominator, and the score is published as not scored rather than as zero.
The Observed Controls Score is calculated across all controls; category scores apply the same formula only to controls in that category. Because unresolved controls leave the calculation, each score is published next to the number of controls it was actually calculated from. Totals are generated from the underlying controls—not entered editorially.
- 1
Observe
Each control in the register is looked for in public sources: contract code, deployment records, audits and documentation.
- 2
Classify
Every control receives one evidence state, and every scored state cites at least one dated source.
- 3
Weight
A control earns its published weight multiplied by the multiplier of its evidence state.
- 4
Score
Points earned divided by applicable weight, times 100, rounded — overall and per category.
- 5
Measure coverage
Separately, the share of in-scope controls carrying a cited source is reported as evidence coverage.
- 6
Decide eligibility
Only protocols with coverage of 80% or more take a numbered position.
03
Worked example
Choose an evidence state for each control in the published register and watch the score, coverage and ranking eligibility respond. The scenarios show how VERIFIED and documentation-supported controls differ, why UNVERIFIED controls are excluded, and why a protocol can score highly and still be provisional.
- Privileged role inventory
- Multisig control of privileged roles
- Pause or freeze capability
- Contract upgradeability
- Upgrade authority
- Timelock on privileged actions
- Mint and burn controls
- Transfer restrictions
- Asset backing and NAV evidence
- Third-party security audits
- Monitoring and security assurance
Result for the hypothetical protocol
Σ(weight × multiplier) = 16
Σ(applicable weight) = 16 (14 excluded)
16 ÷ 16 × 100 = 100 → 100
Provisional — Limited Public Evidence
- A score of 100 with 55% coverage stays provisional: ranking depends on how much of the register is evidenced, not on the score.
- 5 unverified controls are removed from both sides of the formula — the score does not fall, but coverage does.
04
Evidence states
- VERIFIED1×
Direct public evidence — contract source, a deployment manifest or a published audit — supports the control.
- SUPPORTED BY DOCUMENTATION0.75×
Published documentation supports the claim, without direct technical verification.
- NOT OBSERVED0×
The review looked for the control and did not observe it in the cited public scope. This is a finding and scores zero.
- UNVERIFIEDExcluded
Available evidence is insufficient to establish either presence or absence. Excluded from numerator and denominator, and counted against evidence coverage.
- NOT APPLICABLEExcluded
The control does not apply and is excluded from numerator and denominator.
VERIFIED and documentation-supported, side by side
A weight-3 control earns 3 points when VERIFIED and 2.25 when SUPPORTED BY DOCUMENTATION — documentation counts, but less than direct technical evidence. An UNVERIFIED control earns nothing and costs nothing: it leaves the calculation and is reported through evidence coverage instead.
05
Control categories
The register groups its 11 controls into 5 categories. Weights are published per control; the bar shows each category's share of the register's 30 weight points.
Privileged / Admin
5 of 30 weight pointsRole design, access separation, and privileged authority.
Emergency / Upgrade
11 of 30 weight pointsEmergency response, pause, upgradeability, upgrade authority, and timelocks.
Supply / Transfer
6 of 30 weight pointsMint, burn, supply, and transfer restrictions.
Asset Integrity
3 of 30 weight pointsBacking, custody, valuation, and reconciliation signals.
Assurance / Observability
5 of 30 weight pointsAudits, monitoring, disclosures, and operational visibility.
| Control | Category | Weight |
|---|---|---|
| Privileged role inventoryAdministrative and privileged roles are publicly identifiable in contract code or issuer documentation. | Privileged / Admin | 3 |
| Multisig control of privileged rolesPrivileged authority is held by a multi-signature account that can be proven from public sources. | Privileged / Admin | 2 |
| Pause or freeze capabilityA pause, freeze or equivalent emergency stop over the token is observable. | Emergency / Upgrade | 3 |
| Contract upgradeabilityWhether and how the deployed contracts can be upgraded is publicly determinable. | Emergency / Upgrade | 3 |
| Upgrade authorityThe account or contract authorised to perform upgrades is publicly identified. | Emergency / Upgrade | 3 |
| Timelock on privileged actionsA timelock or enforced delay applies to upgrades or other privileged actions. | Emergency / Upgrade | 2 |
| Mint and burn controlsAuthority to mint and burn supply is restricted and publicly documented. | Supply / Transfer | 3 |
| Transfer restrictionsTransfer eligibility is enforced on-chain or through a documented compliance layer. | Supply / Transfer | 3 |
| Asset backing and NAV evidenceBacking of the token and its net asset value are evidenced by published reporting or an on-chain price source. | Asset Integrity | 3 |
| Third-party security auditsAt least one third-party security review of the deployed contract system is publicly available. | Assurance / Observability | 3 |
| Monitoring and security assuranceOngoing monitoring, incident response or an independent controls assurance programme is evidenced. | Assurance / Observability | 2 |
06
Evidence coverage
Evidence coverage is the percentage of in-scope controls with at least one cited source. It is displayed separately and does not add points. A high coverage percentage means more claims are traceable; it does not by itself mean stronger controls. Because an unresolved control carries no source and leaves the score calculation, coverage is where the limits of a review become visible: a high score over few scored controls says less than the same score over a full register.
07
Ranking eligibility
A score says how the observed controls performed; coverage says how much of the register was observable at all. A protocol takes a published position only once its evidence coverage reaches 80%. Below that the score is still published in full, with every control and source behind it, but it holds no position — a high score drawn from a handful of controls is not comparable with the same score drawn from a complete register.
Evidence coverage of 80% or more. The protocol holds a published position in the index.
Evidence coverage below 80%. These scores are calculated from the controls that could be resolved publicly, but evidence coverage is below the threshold required for ranking, so the protocol is published in a separate section and holds no position.
Eligibility is recalculated from the dataset on every build, so a protocol moves between ranked and provisional as its evidence coverage changes — never editorially.
One dataset version
Every protocol in an edition is published against the same dataset version — the snapshot date shown above — and carries it in its own record. Validation rejects a record left on a different version, so no asset is ever compared against a snapshot the rest of the index has moved past.
08
Publication checks
Before publication, automated validation runs against the dataset. Any failure blocks the build.
- Recomputes the Observed Controls Score, every category score, evidence coverage, scored-control counts and ranking eligibility from the controls, and fails on any drift.
- Rebuilds the whole leaderboard, including its order, and fails if the published file differs in any way.
- Confirms the published multipliers and ranking threshold are the ones the code applies.
- Requires a source title, HTTPS URL, observation date and confidence on every scored claim — and no evidence on an unresolved one.
- Requires every protocol to be on the edition's single dataset version, and the live snapshot to match its archive.
- Rejects placeholder or zero contract addresses; an address that cannot be confirmed is published as “Not verified”.
09
Important limitations
- Scores measure publicly observable technical controls and documentation at the stated snapshot date — not controls that exist but are not publicly evidenced.
- Absence of evidence is not evidence of absence. An unresolved control is reported through coverage, never as a failure.
- The index is a point-in-time snapshot, periodically revalidated. It is not continuous monitoring, and deployments can change after the snapshot date.
- A score is not a smart-contract audit, credit rating, investment recommendation, or assessment of an organization's complete security posture.
- A high score drawn from few controls is not comparable with the same score drawn from a full register — which is why coverage decides ranking.