Solution

Transaction confidence scoring

TxnKit helps teams solve confidence scoring with one safe descriptor, confidence-aware output, and warnings when identity evidence is weak.

Ask about a 100-descriptor cleanup auditInspect the API

Target intent

show confidence bands for merchant cleanup decisions.

TxnKit confidence is a decision aid, not a guarantee.

Raw descriptor

POS PURCHASE MAIN ST CAFE

API-shaped output

{
  "normalized_merchant": null,
  "display_name": null,
  "category": null,
  "subcategory": null,
  "website": null,
  "logos": {
    "preferred": null,
    "variants": [],
    "warnings": [
      "no_verified_logo_for_low_confidence_descriptor"
    ]
  },
  "confidence": 0.25,
  "recurring_hint": false,
  "processor_hint": "unknown",
  "location_hint": null,
  "signals": [
    "generic_card_descriptor"
  ],
  "warnings": [
    "weak_merchant_evidence"
  ]
}

Low-confidence results should keep fallback labels and warnings visible instead of forcing a guessed logo or website.

When to use

  • You need confidence scoring for transaction-feed UI, audit, or import review workflows.
  • You can send safe anonymized descriptors without customer PII or full statement text.
  • You want API-shaped output that a developer can inspect before wiring product UI.

When not to use

  • TxnKit confidence is a decision aid, not a guarantee.
  • Do not use TxnKit to process card numbers, account numbers, full statements, bank credentials, emails, phone numbers, addresses, customer names, or customer PII.
  • Do not use TxnKit when the product requirement is guaranteed merchant identity, every-merchant resolution, or contracted uptime terms.

Proof surfaces

The public contract is POST /v1/enrich, backed by the OpenAPI file, benchmark examples, privacy rules, and deterministic request-path tests.

OpenAPI · Benchmark · Security · Pricing

Related pages