Data Architecture

Data Domains

Data is organised into clearly defined domains owned by the corresponding services. Cross-service access occurs through APIs or event streams, not through shared database access.

Domain Examples of Data Retention Driver
Identity User profile, KYC attributes, device bindings, authentication events. Regulatory (KYC/AML retention rules per jurisdiction).
Ledger Transactions, entries, balances, reconciliations. Statutory accounting and audit requirements.
Payments Payment requests, rail selections, partner references, callbacks. Rail and partner dispute windows; AML record-keeping.
Cards Card metadata, tokens, authorisations, disputes. Scheme (Visa/Mastercard) and PCI DSS requirements.
Credit Credit builder loan terms, repayment schedules, bureau reports. FCRA and bureau retention rules.
Compliance Screening hits, case files, SAR/STR records. Regulatory retention (typically 5+ years).
Analytics Aggregated, de-identified usage and performance data. Internal analytical use.

Sensitive Data Classification

Data is classified into four tiers, each with distinct handling requirements:

Tier Description
Tier 1 — Regulated cardholder data Full PANs, CVV, track data. NEVER stored in Kredete systems; handled only by PCI-DSS-compliant partner processors. Systems referencing these values handle tokens only.
Tier 2 — Personally identifiable information (PII) Names, addresses, dates of birth, government ID numbers, KYC documents, biometric templates. Encrypted at rest and in transit; access is role-based and fully audited.
Tier 3 — Financial and transactional data Balances, transactions, ledger entries, partner references. Encrypted at rest; strict service-to-service access controls.
Tier 4 — Operational and analytical data De-identified, aggregated, or non-sensitive attributes used for product analytics and reporting.

Sensitive Data Flow

Sensitive data traverses the platform according to the following principles:

  • All client–server traffic is TLS 1.2+ encrypted; internal service-to-service traffic is mutually authenticated and encrypted.
  • Card data is captured directly into partner-hosted fields or SDKs, never passing through Kredete application servers in the clear.
  • KYC documents are uploaded directly to object storage using signed, short-lived upload URLs; server-side encryption is applied on write.
  • Biometric templates used for face authentication are stored as irreversible representations; raw images are not retained beyond the verification step.
  • Database-level encryption uses managed keys under the organisation's key-management service, with key rotation performed on a defined schedule.

Data Storage Components

  • Primary transactional stores: relational databases (e.g., PostgreSQL) for ledger, accounts, payments, and compliance domains, deployed as managed clusters with synchronous replication and point-in-time recovery.
  • Caching and in-memory stores: managed Redis for session state, rate limiting, FX quote caching, and hot read paths.
  • Event streaming: a managed message bus used for domain events, audit log ingestion, and asynchronous orchestration between services.
  • Object storage: used for KYC documents, exported reports, and other large artefacts, with server-side encryption and object versioning enabled.
  • Analytical store: a columnar data warehouse for business intelligence, product analytics, and regulatory reporting, fed from the operational stores via change-data-capture and event streams.