I spent two weeks last winter rewriting the procurement narrative our infrastructure team had been telling itself for nearly four years. The pitch had always been the same: stitch together six fast-moving open-source projects, hire a platform team to keep them aligned, and call the result a competitive advantage. By the third on-call rotation of the quarter, nobody on the staff engineering bench believed it anymore.
What the audit actually showed
When the controller pulled the run-rate, two-thirds of our cloud spend was sitting in a query layer none of the analysts trusted. The schema had drifted in eleven places. The runbook was a private Notion page. We could not, on a Tuesday morning, answer the question of which row in the ledger had last been touched, or by what.
The CFO did not ask for a roadmap. She asked which database had been answering that question, correctly, for the last forty years.
That afternoon we drafted a one-page memo. The recommendation was unglamorous: consolidate on a single relational engine with a published support contract, hand the operational pager to a vendor whose business model is keeping it quiet, and reclaim the four engineers we had stationed full-time on yak-shaving.