Ingestion batches preserve analytics lineage

What the public implementation shows

Guevara models transaction and fraud ingestion batches and exposes a GetTransactionIngestionBatchesQuery alongside webhook commands.

Webhook reception is recorded as a batch-level operation while individual transaction records remain queryable. This is the implementation fact behind the note; the architectural reading begins from that visible boundary rather than from an assumed production story.

Operating principleAnalytics is trustworthy only when the path from source batch to derived view can be reconstructed.

The decision behind the mechanism

Batch identity makes partial ingestion, replay, source timing, and derived analytics runs explainable.

Rows without an ingestion envelope cannot show whether absence means no source activity, a failed batch, or an incomplete delivery. The mechanism matters because it gives that failure a named boundary, durable evidence, and a controlled response instead of leaving the outcome to timing or operator guesswork.

• Assign a durable identity to every incoming batch.
• Record accepted and rejected row counts.
• Make replay behavior idempotent.
• Expose batch freshness with analytical views.

How the control should be operated

A batch record should carry source, received time, expected and accepted counts, rejection counts, processing status, and replay relationship.

Analytics consumers can assess freshness and completeness before acting on the data. The reusable lesson is not to copy a class or endpoint in isolation, but to preserve the relationship between identity, state, authority, evidence, and operational ownership.