Blog

Technical notes on transaction enrichment.

Concrete examples for builders who need merchant names, confidence, warnings, logo metadata, and privacy boundaries without turning the request path into live research.

August 14, 2026 How to parse processor-wrapped transaction descriptors Separate processor wrappers, merchant phrases, store tokens, and location hints before making a merchant identity claim. July 27, 2026 What a transaction enrichment API should return on a miss Return a readable fallback, confidence, signals, and warnings while omitting merchant metadata the evidence cannot support. July 8, 2026 Normalization vs. merchant identification: why the distinction matters Clean descriptor text is useful, but it is not the same as a reviewed merchant identity, website, or logo. June 20, 2026 Why one-descriptor API requests are easier to trust than ambiguous batches Keep validation, timeouts, confidence, warnings, retries, and failures attributable to one transaction at a time. June 1, 2026 How to test merchant enrichment with deterministic fixtures Test recognized, unresolved, and privacy-rejected cases with evidence-aware assertions instead of display-name checks alone. May 18, 2026 Why descriptor cleanup should return warnings, not fake certainty Confidence-aware enrichment lets product teams design fallback UI instead of trusting a guessed merchant identity. May 4, 2026 How to enrich transactions without sending sensitive financial data Reject statement-like and PII-heavy input before parsing, logging, matching, queueing, or third-party use. April 17, 2026 MCC is useful context. It is not merchant enrichment. MCC can help category confidence, but it cannot turn processor-wrapped descriptors into display-ready merchants by itself. April 5, 2026 Why TxnKit avoids live LLM calls during POST /v1/enrich A transaction row should not wait on a prompt, crawler, logo lookup, or third-party enrichment provider. March 28, 2026 Logo metadata belongs in transaction enrichment, not in a separate guessing step Logo variants need source, purpose, shape, size, theme, and confidence fields so client apps can choose safely.