How the research is made
Every venue gets a written, dated v1 memo with observable review conditions: the specific events that would reopen the assessment. Retired records move into the dated archive rather than disappearing; the audit trail is the product. Category rules may set firm shelf policy, but they do not masquerade as protocol-quality ratings. Favorable research requires an individual review.
Ketju checks four parts of a position: the asset (who can freeze it), the protocol (what its governance can do), the chain (who orders transactions and holds upgrade keys), and the custody arrangement (who holds the keys). A sound protocol cannot remove control held by an issuer, chain operator, or custodian. Source TVL is reported as scale context only. Executable withdrawal capacity requires a protocol-specific, proposed-size reading.
Atlas comparisons follow policy v2.0. A peer must use the same kind of asset for the same kind of client purpose, and its protocol must be shelf-eligible. A stablecoin is never replaced with ETH to improve a control result. The closest structure appears first, followed by reported TVL only as a scale tie-breaker. Yield is shown but never selects or ranks a peer. When no position passes those rules, the Atlas says so.
Size is an intake and capacity signal, never a quality verdict. A smaller venue can remain capacity-unproven without being called unsafe. Before implementation, the advisor must obtain an executable exit reading for the proposed amount and record any uncertainty explicitly.
Research is not an annual filing. The Board carries live venue numbers, alerts are checked every day, and shelf-eligible records are re-examined at least quarterly. A review condition reopens the memo that week, not at the next scheduled date. Adverse, unresolved, and policy-excluded records remain in the research universe on an evidence-appropriate cadence.
Ketju accepts no compensation from any protocol, chain, or issuer it covers. Observed yield is descriptive and never selects, ranks, approves, or rejects a position.
Research governance publishes the publication, correction, source-receipt, cadence, and incident policies, with the current split coverage and freshness measures.