Situation before the analytics reset
- Critical metrics lived in overlapping warehouses, exports, and local models with weak shared definitions.
- Regulatory and client pressure required faster answers without weakening access control or audit evidence.
- The scope focused on governed delivery and operating cadence, not a vague rip and replace of every legacy store.
Data quality and trust
Core points
- Duplicate definitions meant two teams could be right in isolation yet wrong for the firm.
- Late discovery of quality issues embarrassed client conversations that should have been routine.
- Without explicit owners, cleansing work fell to whoever had time after quarter close.
Architecture and platform choices
Core points
- Ingestion pipelines had to tolerate bursty loads while staying observable enough to debug in hours.
- Sensitive advisory datasets required least privilege paths, not broad shared drives with best effort masking.
- Tool sprawl was trimmed only when integration contracts and exit criteria were written before purchase.
Delivery milestones
Core points
- Phased domains landed value early instead of freezing the organisation on a multi-year programme.
- Automated tests around pipelines and transforms caught regressions where humans used to spot check.
- Training sat beside rollout so self-service tools did not become shadow IT with prettier buttons.
Skunk tip
- Publish a one page RACI for metric ownership before you debate another warehouse schema.
Measured outcomes and lessons
Core points
- Where leadership published variance targets, teams improved cycle time on trusted reports measurably.
- The hardest lesson was that catalog discipline beats another headline feature if nobody maintains it.
- Continuous improvement replaced a big bang finish line, which kept funding aligned with risk reduction.
If your executives still ask which spreadsheet is canonical, you have not finished the governance conversation.
