Research data points and field semantics

AMLOK’s research representation connects public financing evidence to financial activity without treating every source as a complete borrower record. You can use this dictionary to identify what a source directly supplies, what requires an additional source, and which calculations remain unavailable.

Evidence envelope

Every extracted value carries its original representation and a typed normalized representation. A decimal without its currency, scale, period and meaning is not a complete financial observation.

Field group	Required data points	Validation rule
Provenance	Source ID, publisher, URL, file digest, acquisition timestamp, query/release version	A claim must resolve to the acquired version, not just a homepage.
Locator	PDF page and printed page, table/row/column or text offsets	Keep physical PDF numbering distinct from printed numbering.
Time	Event time, publication time, first verified availability, observation period, retrieval time	Never substitute retrieval time for the historical event date.
Interpretation	Raw text, normalized value, units, currency, range bounds, status	Preserve source precision and conditions.
Extraction	Method, model/version, confidence, reviewer state, correction lineage	Confidence is a prediction, not independent evidence.
Rights	Allowed access, evaluation/training use, redistribution and retention	Public visibility alone does not grant all uses.
Missingness	Not disclosed, not applicable, unavailable, redacted, ambiguous or extraction failed	Missing is not zero and is not a negative outcome.

Identity and financing

Data point	Direct evidence route	Interpretation
Legal name and identifier	Corporate registries; identifiers in executed filings	Preserve jurisdiction and effective name history.
Borrower, issuer and subsidiary	Agreement parties and issuer disclosures	Do not assign a subsidiary obligation to every group company.
Lender, agent and guarantor roles	Financing agreements, EDC records, security registrations	An agent is not necessarily the only lender.
Country and province	Source jurisdiction and explicitly disclosed location	Export destination, borrower location and exporter location differ.
Industry	Published sector or NAICS classification	Preserve classification version and source granularity.
Financing product	Loan, revolver, guarantee, equity, quasi-equity or other stated type	A guarantee limit is not the borrower’s cash inflow.
Signing, close and disbursement dates	Explicit event statement or authoritative dated field	Keep distinct even when dates coincide.
Commitment amount	Stated contractual limit	Not equal to funded principal or present debt.
Drawn amount and outstanding balance	Dated financial disclosure or authorized account evidence	Drawn at a reporting date may differ from original proceeds.
Available capacity	Explicit table field or documented calculation	Keep letters of credit and commercial-paper backing separate.
Amount range	Disclosed lower/upper bounds and boundary inclusivity	Do not replace a band with its midpoint and call it observed.
Currency and magnitude	Table heading, footnote, agreement or source dictionary	CAD millions and USD units must not be added without conversion.
Interest basis	Fixed/variable indicator, base rate, spread and step conditions	A spread is not an all-in interest rate.
Term and maturity	Effective term, maturity date, extension option and conditions	Options are not exercised extensions.
Amortization and repayment	Explicit schedule, mandatory payments and prepayment terms	Unknown repayment terms remain unknown.
Covenants and security	Exact clause, measurement period, threshold and collateral description	Do not infer breach from a generic covenant reference.
Status and amendments	Executed, conditional, renewed, amended, repaid or source-specific state	Preserve both the source state and normalized event.

These fields are supported by the source register and worked public records. Their presence varies by source and record; the dictionary is not a claim of universal coverage.

What public records can establish

Evidence	Available research labels	Labels it does not establish alone
Corporate registration	Identity attributes and effective registry status	Income, creditworthiness or debt repayment.
Security registration	Named debtor, secured-party relationship and registered collateral	Cash advanced, current balance, default or a bank debit descriptor.
Signed financing disclosure	Stated parties, product, date, amount or band	Complete funding path or subsequent payment performance.
Financial statements	Reported period balances, flows and selected debt terms	Individual account transactions or all unreported obligations.
Court or insolvency record	Precisely identified procedural event or adjudicated finding	Every allegation as fact, or every filing as a contractual default.
Published program aggregates	Cohort totals and historical program context	Identifiable loan examples or individual outcomes.

Permissioned activity needed for transaction understanding

Public financing records form the corroboration and document corpus. The following record families require actual authorized activity data for a linked, multi-source borrower benchmark. Do not fabricate a bank ledger by expanding a published annual total.

Record family	Research data points	Linkage question
Bank transaction	Account reference, posting/value date, direction, amount, currency, descriptor, pending/posted status and available balance	Is the inflow revenue, financing, transfer, refund or another event?
Payment processor	Charge/payment ID, gross amount, fee, refund, dispute, payout ID and settlement time	Which sales and adjustments produced a net payout?
Commerce	Order, line items, discounts, tax, shipping, fulfillment, cancellation and return	Which commercial activity precedes or reverses a payment?
Accounting	Journal, ledger account, invoice, bill, credit note, receivable/payable status and period	Is cash movement consistent with accrual accounting?
Payroll and consumer income	Pay period, gross/net payment, source identity, deductions and recurrence	Is income recurrent and attributable without inventing employer identity?
Lending evidence	Facility, installment schedule, paid amounts, days past due and authorized outcome	What is owed, what was paid, and what outcome definition applies?

Record identifiers alone are not cross-provider joins. Test joins against shared references, amounts, currency, timing, netting and independently reviewed relationships. One payment can settle several invoices, and one bank deposit can aggregate many payments.

Financial features and admissible inputs

Feature family	Required inputs	Calculation constraint
Operating cash flow	Classified bank activity and linked business records	Exclude transfers and financing from revenue; report unclassified coverage.
Revenue quality	Orders, payments, refunds, disputes and settlement links	Separate gross sales, net sales and net settled receipts.
Liquidity	Available cash, supported obligations and time windows	Distinguish observed liquidity from an assumption-dependent projection.
DSCR	Defined cash flow available for debt service and contractual debt service over the same period	Show numerator, denominator, adjustments and missing debt coverage.
Consumer DTI	Eligible income and debt obligations under a stated definition	Keep gross/net income conventions and monthly units explicit.
Concentration	Resolved customers, suppliers, funders and attributable volumes	Report unresolved exposure instead of forcing a counterparty match.
Working capital	Receivables, payables, inventory and relevant flows	Align stock dates and flow periods; avoid double counting invoice payments.
Stability and seasonality	Sufficient dated observations and coverage	Distinguish seasonality from missing feeds and one-time events.
Monitoring change	Versioned baseline, newly available events and materiality rules	A data correction is distinguishable from a new economic event.

The financial analysis specification defines formulas and the record model defines event reconciliation. Derived features retain source references and calculation versions so a reviewer can reconstruct them.

Amounts, dates and ambiguity

Store decimal quantities without binary floating-point conversion at ingestion. Preserve source rounding. A printed dash is interpreted only after checking the table convention; an empty cell stays absent. Use interval-valued amounts when a source supplies a band. Record date precision as day, month, quarter or year rather than inventing a more precise date. For a conditional facility, preserve initial availability, conditional capacity and total potential commitment as separate observations. For bilingual names, retain the original accents and legal form; use a normalized matching field in addition to, not in place of, the source name. Every derived relationship identifies whether it is directly stated, corroborated or still a candidate.

Research program

Data and models

Risk analysis

Evaluation and evidence

Research data points and field semantics

Evidence envelope

Identity and financing

What public records can establish

Permissioned activity needed for transaction understanding

Financial features and admissible inputs

Amounts, dates and ambiguity

​Evidence envelope

​Identity and financing

​What public records can establish

​Permissioned activity needed for transaction understanding

​Financial features and admissible inputs

​Amounts, dates and ambiguity

Evidence envelope

Identity and financing

What public records can establish

Permissioned activity needed for transaction understanding

Financial features and admissible inputs

Amounts, dates and ambiguity