Screening · adjudication

Match scorecards

A scorecard turns raw match scores into a decision. Match settings (on the screening console) decide which entities are candidates; a scorecard decides how to treat each candidate — a band and a recommended disposition. Build one here, test it against the public index, then pick it on the console. Everything is saved only in this browser.

How scorecards work

Bands vs dispositions. A band is a risk tier from the score (cleared / review / match); a disposition is the recommended action.
Block vs reject are legally distinct. Block holds property (OFAC 10-day clock); reject holds nothing — it refuses or declines.
Test before you rely on it. A card can’t be activated on the console until you run at least one Test run (governance-by-evidence — there is no second approver in a single browser).
Greyed criteria are part of the full model but not yet in the public index (e.g. government ID, alias strength), so they’re shown but disabled.
Hand-off. Exported JSON is byte-identical to what the admin control-plane and the offline appliance consume — a public draft lifts straight into a governed deployment.

Your scorecards

No scorecards yet — start from a template or a blank card below.

Start from a template

blankBlank scorecardA minimal valid card — a type-mismatch veto, a name+DOB baseline, and three bands. Build up from scratch.
global_bankGlobal / Tier-1 BankStrictest calibration: hard vetoes on type / ID / DOB conflict, a weak-AKA name discount, and a thin-record review floor. Default-to-queue bands.
Plain-language rules
Applies when topics {sanction} and schema {Person}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Reject the match (auto-clear as a non-match) if the government ID / passport conflicts. Reject the match (auto-clear as a non-match) if the date of birth conflicts. Raise the result to at least the 'review' band if the name score is at least 90 and the government ID / passport is missing and the date of birth is missing. When only a weak (LowQuality) alias matched, scale name by ×0.5. Otherwise score name (50%) + ID (45%) + DOB/country (45%) [-0.7 if it conflicts] + nationality (10%) + address (8%); below 0.72 → Auto-clear (audit record), 0.72–0.88 → Route to L1 review, 0.88–0.95 → Route to L2 review, 0.95 and above → Escalate to MLRO.
mid_size_bankRegional BankNear-default with a soft DOB conflict penalty and a review floor — regional-bank posture.
Plain-language rules
Applies when schema {Person}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). When only a weak (LowQuality) alias matched, scale name by ×0.6. Otherwise score name (55%) + ID (45%) + DOB/country (42%) [-0.7 if it conflicts] + nationality (10%); below 0.58 → Auto-clear (audit record), 0.58–0.85 → Route to L1 review, 0.85 and above → Escalate to MLRO.
fintech_paymentsFintech / PaymentsAggressive straight-through auto-clear (high review floor) with hard conflict vetoes; no thin-record floor.
Plain-language rules
Applies when screening context {payment, onboarding}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Reject the match (auto-clear as a non-match) if the government ID / passport conflicts. Reject the match (auto-clear as a non-match) if the date of birth conflicts. When only a weak (LowQuality) alias matched, scale name by ×0.5. Otherwise score name (55%) + ID (45%) + DOB/country (42%) [-0.7 if it conflicts]; below 0.78 → Auto-clear (audit record), 0.78–0.90 → Route to L1 review, 0.90–0.95 → Escalate to MLRO, 0.95 and above → Hold / BLOCK.
crypto_vaspCrypto / VASPIdentifier / wallet-weighted with an ID-conflict veto — VASP posture.
Plain-language rules
Applies when topics {sanction, crime} and schema {Person, Organization}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Reject the match (auto-clear as a non-match) if the government ID / passport conflicts. Otherwise score name (45%) + ID (55%) + DOB/country (30%) [-0.7 if it conflicts]; below 0.70 → Auto-clear (audit record), 0.70–0.85 → Route to L1 review, 0.85–0.93 → Route to L2 review, 0.93 and above → Escalate to MLRO.
insurerInsurerLifecycle-scoped; a strong hit routes to hold / block.
Plain-language rules
Applies when screening context {onboarding, periodic_rescreen, claim}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Otherwise score name (55%) + ID (45%) + DOB/country (42%) [-0.7 if it conflicts] + nationality (10%); below 0.62 → Auto-clear (audit record), 0.62–0.85 → Route to L1 review, 0.85 and above → Hold / BLOCK.
export_controls_corporateExport-Controls CorporateWidest dataset scope, low review floor, thin-record floor — export-controls corporate posture.
Plain-language rules
Applies when datasets {us_ofac_sdn, us_bis_dpl, us_bis_entity, eu_fsf, gb_hmt} and schema {Person, Organization}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Raise the result to at least the 'review' band if the name score is at least 85 and the government ID / passport is missing and the date of birth is missing. Otherwise score name (55%) + ID (45%) + DOB/country (30%) [-0.7 if it conflicts] + address (10%); below 0.50 → Auto-clear (audit record), 0.50–0.80 → Route to L1 review, 0.80–0.92 → Route to L2 review, 0.92 and above → Reject transaction.
defaultUniversal defaultUniversal fallback — reproduces the screening baseline exactly. A customer who never opens the builder sees identical behaviour.
Plain-language rules
Applies to every candidate. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Otherwise score name (55%) + ID (45%) + DOB/country (42%) [-0.7 if it conflicts]; below 0.60 → Auto-clear (audit record), 0.60–0.85 → Route to L1 review, 0.85 and above → Escalate to MLRO.

Import a scorecard

Paste a scorecard JSON (kernel doc). It is validated through the same kernel the appliance uses before it is accepted.

Loading your local scorecards…

Match scorecards

Start from a template

blankBlank scorecardA minimal valid card — a type-mismatch veto, a name+DOB baseline, and three bands. Build up from scratch.

global_bankGlobal / Tier-1 BankStrictest calibration: hard vetoes on type / ID / DOB conflict, a weak-AKA name discount, and a thin-record review floor. Default-to-queue bands.

Plain-language rules

Applies when topics {sanction} and schema {Person}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Reject the match (auto-clear as a non-match) if the government ID / passport conflicts. Reject the match (auto-clear as a non-match) if the date of birth conflicts. Raise the result to at least the 'review' band if the name score is at least 90 and the government ID / passport is missing and the date of birth is missing. When only a weak (LowQuality) alias matched, scale name by ×0.5. Otherwise score name (50%) + ID (45%) + DOB/country (45%) [-0.7 if it conflicts] + nationality (10%) + address (8%); below 0.72 → Auto-clear (audit record), 0.72–0.88 → Route to L1 review, 0.88–0.95 → Route to L2 review, 0.95 and above → Escalate to MLRO.

mid_size_bankRegional BankNear-default with a soft DOB conflict penalty and a review floor — regional-bank posture.

Plain-language rules

Applies when schema {Person}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). When only a weak (LowQuality) alias matched, scale name by ×0.6. Otherwise score name (55%) + ID (45%) + DOB/country (42%) [-0.7 if it conflicts] + nationality (10%); below 0.58 → Auto-clear (audit record), 0.58–0.85 → Route to L1 review, 0.85 and above → Escalate to MLRO.

fintech_paymentsFintech / PaymentsAggressive straight-through auto-clear (high review floor) with hard conflict vetoes; no thin-record floor.

Plain-language rules

Applies when screening context {payment, onboarding}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Reject the match (auto-clear as a non-match) if the government ID / passport conflicts. Reject the match (auto-clear as a non-match) if the date of birth conflicts. When only a weak (LowQuality) alias matched, scale name by ×0.5. Otherwise score name (55%) + ID (45%) + DOB/country (42%) [-0.7 if it conflicts]; below 0.78 → Auto-clear (audit record), 0.78–0.90 → Route to L1 review, 0.90–0.95 → Escalate to MLRO, 0.95 and above → Hold / BLOCK.

crypto_vaspCrypto / VASPIdentifier / wallet-weighted with an ID-conflict veto — VASP posture.

Plain-language rules

Applies when topics {sanction, crime} and schema {Person, Organization}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Reject the match (auto-clear as a non-match) if the government ID / passport conflicts. Otherwise score name (45%) + ID (55%) + DOB/country (30%) [-0.7 if it conflicts]; below 0.70 → Auto-clear (audit record), 0.70–0.85 → Route to L1 review, 0.85–0.93 → Route to L2 review, 0.93 and above → Escalate to MLRO.

insurerInsurerLifecycle-scoped; a strong hit routes to hold / block.

Plain-language rules

Applies when screening context {onboarding, periodic_rescreen, claim}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Otherwise score name (55%) + ID (45%) + DOB/country (42%) [-0.7 if it conflicts] + nationality (10%); below 0.62 → Auto-clear (audit record), 0.62–0.85 → Route to L1 review, 0.85 and above → Hold / BLOCK.

export_controls_corporateExport-Controls CorporateWidest dataset scope, low review floor, thin-record floor — export-controls corporate posture.

Plain-language rules

Applies when datasets {us_ofac_sdn, us_bis_dpl, us_bis_entity, eu_fsf, gb_hmt} and schema {Person, Organization}. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Raise the result to at least the 'review' band if the name score is at least 85 and the government ID / passport is missing and the date of birth is missing. Otherwise score name (55%) + ID (45%) + DOB/country (30%) [-0.7 if it conflicts] + address (10%); below 0.50 → Auto-clear (audit record), 0.50–0.80 → Route to L1 review, 0.80–0.92 → Route to L2 review, 0.92 and above → Reject transaction.

defaultUniversal defaultUniversal fallback — reproduces the screening baseline exactly. A customer who never opens the builder sees identical behaviour.

Plain-language rules

Applies to every candidate. Reject the match (auto-clear as a non-match) if the query and candidate are different entity types (Person vs Organization). Otherwise score name (55%) + ID (45%) + DOB/country (42%) [-0.7 if it conflicts]; below 0.60 → Auto-clear (audit record), 0.60–0.85 → Route to L1 review, 0.85 and above → Escalate to MLRO.