A mid-size manufacturing supplier watched their Looker dashboards crumble after a dataset migration. In 48 hours, Ironhide found the failure points, fixed the joins and refresh chain, and restored trust in the numbers.
Two days. Zero drama. Dashboards they can actually run the week from.
After an internal migration, dashboards started failing: broken joins, mismatched time zones, and totals that didn’t match the ERP. The ops team lost confidence and slipped back to manual Excel exports.
They didn’t need a rebuild. They needed a surgical rescue: diagnose the failure points, fix the math, and harden the refresh chain — all without blowing up the week.
We traced the failure path from source to dashboard, corrected the schema and time handling, and rebuilt the most brittle transforms in SQL. Then we verified totals against the ERP and restored notifications so problems surface immediately.
Five frames that tell the story: the restored dashboard, the before→after, the control room, the SQL fix, and the team back in sync.
Repaired model references and simplified LookML so visuals load quickly and reflect the corrected data model.
Moved heavy transforms to SQL, fixed joins and timestamp casts, and added reconciliation queries to validate totals against ERP exports.
Used prior “last trusted” exports to confirm KPI math before turning dashboards back on.
Re-enabled notifications so refresh failures or data drift trigger immediate action — not a week of bad decisions.
Ground truth for jobs, runtime, and postings; the source we reconcile to so Finance and Ops stay aligned.