> ## Documentation Index
> Fetch the complete documentation index at: https://docs.amlok.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Normalization rules

> Rules for evidence, missing values, money, identity, and normalized data.

The normalization rules below are drawn from the workbook’s **Normalization & API Rules** tab. HTTP and webhook specifications are documented in the [API reference](https://docs.amlok.io/api-reference/overview.md).

## Version

This data reference accompanies AMLOK API **1.4.2**. Review [versioning](https://docs.amlok.io/api-reference/versioning.md) when adopting changes to schemas or calculations.

## Scope

Normalized union of credit-relevant fields across 238 inventory entries, plus AMLOK verification services, AMLOK registry services and PPSA Canada. Provider-specific extras use documented namespaced extensions.

This is not an exhaustive dump of every vendor API field.

## Field evidence

Upstream field references identify source evidence. Resource mappings describe how provider data relates to AMLOK resources; field availability and permitted use depend on the connected product and agreement.

See Field Mappings and Provider Coverage.

## Required vs available

Required describes a valid AMLOK resource when that resource exists. It does not guarantee a source supplies every optional field or that every resource exists.

Required source IDs must be resolved or the record is quarantined.

## Shared metadata

Every top-level resource inherits common fields. Resource-name prefixes in Field Dictionary are documentation context; response data contains the relative field names.

Shared Types expands every reusable nested object.

## Null semantics

Use null for known missing/unavailable optional values and include missing\_fields reasons. Zero and false require evidence. An empty array means the requested collection was queried and is known empty.

Denied sensitive fields are omitted; missing\_fields may explain redaction without exposing values.

## Time

Dates preserve their source calendar date. Timestamps use RFC 3339 UTC with original timezone in provenance where needed. Date-only evidence is never padded into an invented timestamp.

posted\_on, observed\_at, effective\_at and retrieved\_at remain distinct.

## Money

Money uses exact decimal strings in major currency units and a currency code. Convert minor-unit sources using currency-specific exponent tables. Preserve higher precision when source requires it.

No binary floating-point financial calculations; do not assume all currencies have two decimals.

## Signs

Bank transaction amount is an absolute magnitude with separate direction. Processor ledger net and gross can be signed. Loan principal is an obligation magnitude.

Convert per provider and product; do not use one global sign rule.

## Ratios

Rates and proportions are decimal fractions: 9.5% is 0.095. DSCR and current ratio are multiples, not percentages.

Units and denominator policies are required.

## Currency conversion

Keep original money. Cross-currency aggregates require an explicit FX source, effective time, rate convention and rounding method.

Never sum CAD and USD without stated conversion.

## Identifiers

AMLOK IDs are opaque and tenant-scoped. Provider identifiers remain case-sensitive strings qualified by provider, connection and resource type.

Preserve leading zeroes; masks and names are not keys.

## Name and address normalization

Preserve original spelling, scripts and address text. Store matching normalization as a separate versioned transformation.

No automatic merge on fuzzy name similarity alone.

## Country and industry codes

Normalize countries to alpha-2 where verified. Preserve jurisdiction, source code and taxonomy version for registry and industry fields.

NAICS, SIC, MCC and NACE are not interchangeable.

## Entity resolution

Relationships require evidence. Keep unresolved counterparties and review state rather than forcing a legal-entity match.

Canonical merging must be reversible and audited.

## Deduplication

Deduplicate within a source using stable keys. Link cross-source order, invoice, payment, payout and bank records with reconciliation evidence.

These records describe different economic stages and must not be summed as independent revenue.

## Source precedence

Use purpose-specific authoritative sources, observation dates and review state. Preserve conflicting values in provenance.

A registry source may establish formation; a borrower assertion cannot silently overwrite it as verified.

## Changes and history

Use source cursors/webhooks where available. Store record versions, pending-to-posted linkage, removals and immutable assessment input manifests.

AMLOK change cursors are separate from provider cursors.

## Coverage

Return actual time span, sources, account set and jurisdictions completed. Failed or unqueried sources remain explicit.

Partial success never becomes a completed no-hit result.

## Freshness

Provider update frequency, AMLOK polling, ingestion time and source effective time are distinct.

No universal real-time or historical-depth promise.

## Sensitive data

Restricted fields use explicit entitlements and purpose. Default identifiers are masked or tokenized. Raw biometric media, PAN, CVV and provider credentials are excluded from the normalized default response.

Per-client retention, residency and access policies must be implemented.

## Derived attributes

Every computed value records inputs, window, unit, calculation/model version and evidence coverage. Missing prerequisites produce unavailable status.

No invented credit score, PD, fraud probability or approved limit.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.