Coldpipe

Small companies sometimes inherit a cloud warehouse after a consultant or analyst leaves, with scheduled queries, tables and dashboards that nobody fully owns. Coldpipe proposes a read-only janitor: observe job and query history, build a dependency candidate graph, identify recurring cost, flag pipelines that stopped producing expected data and prepare a monthly reclaim plan. The supplied research confirms engineer-focused optimization products and no known self-serve product for the engineer-less small-business segment, though direct customer pain evidence is absent.

Observed query history is incomplete lineage. A table with no recorded reader may support an external export, manual process, regulatory retention or infrequent year-end report. Source metadata, query event, parsed dependency, inferred owner, cost attribution, freshness expectation, unused candidate, failure candidate, business-owner disposition, change approval, warehouse command, acknowledgement, readback and realized savings must remain separate. Coldpipe must never delete, pause or rewrite production in its initial mode.

The map can become valuable over time, but customer export and verification should prevent lock-in through obscurity. Numeric warehouse and competitor prices are omitted because they are observed market references, not fixed product pricing.

An operations, finance or analytics owner at a small company that pays for a warehouse but lacks a dedicated data engineer.