When two dashboards tell two different stories

A car manufacturer ran GA4 across fourteen national websites, and the numbers never agreed with the sales system behind them. In a bad month the same figure could differ by a fifth depending on which tool you asked. People had quietly stopped opening the dashboards, which meant the data was informing nothing.

The problem lived at collection, not in the reporting. Tag audits and checks in DebugView showed the tracking had drifted market by market: the same "lead" event fired on three different triggers, and a scope error was double-counting sessions on two of the countries. Close to a fifth of the entire gap traced back to a single misconfigured form on the highest-traffic site, a small collection error with an outsized reporting cost.

The fix was structural. One shared event specification was rebuilt and applied across all fourteen properties, GA4 was reconciled against the CRM one market at a time to isolate exactly where each number broke, and the tagging standard was documented so the drift could not quietly return.

Once collection was clean, the two systems landed within 3% of each other, and marketing went back to allocating budget from the dashboard instead of guessing. A number nobody trusts is worse than no number at all, and that trust is built at collection, never in the report.