One group. Four systems. No single truth.

A multi-rooftop automotive dealership group ran on four disconnected vendor systems — dealer management reporting, CRM, inventory and market pricing, web and ad analytics. None of them agreed. Two products from the same vendor reported different pictures of the lot, and the numbers leadership needed most were missing or untrustworthy. Every decision opened with an argument about whose report was right.

The vendors don't make this easy. No clean APIs — data arrives as scheduled file drops and rate-limited exports, on the vendors' terms. Critical fields are withheld from standard exports entirely. And the application this foundation powers is proprietary, so the layer had to be built without waiting on any vendor's roadmap.

A vendor-neutral data layer: hardened, firewalled ingestion with jailed one-way vendor connections; every feed normalized into one governed schema, source-tagged and joined on shared keys; raw payloads preserved immutably so any load can be replayed and audited; and a governed metric layer where every number is defined once. The AI reasons over pre-computed, governed metrics — never freelance queries against raw financial tables.

Leadership starts from numbers everyone trusts. The arguments moved from “whose report is right” to “what do we do about it.” The engagement is ongoing — the foundation absorbs new sources as the group grows.