When data discrepancies start showing up across teams, the default response is usually the same. We need better governance. Governance matters, but in many cases governance is not the real starting point. It is an operating model problem disguised as governance. What usually sits underneath is much more practical and much more uncomfortable: no clear owner of metric logic, unclear transformation steps from raw data to reported number, different mapping tables and aggregation rules across teams, and a general reluctance to revisit existing reports because doing so may expose historical inconsistency. In many cases, the hardest part is that nobody really wants to own the messy reconciliation work.
That is why two teams can start from what appears to be the same raw data and still end up with different answers. The issue is rarely just the number itself. It is the hidden assumptions between source and output. Different extracts, different refresh timing, different filters, different aggregation grain, different mapping logic, and different handling of duplicates, nulls, transfers, cancellations, or late arriving records all create divergence long before the number reaches a dashboard. When those conditions exist, teams are not really debating one number. They are debating every undocumented assumption that sits behind it.
I suppose..
A lot of what we call governance failure is actually a failure in ownership, decision rights, and accountability. Governance can help define rules, assign owners, and create change control, but governance alone cannot reconcile the numbers, clean up messy historical logic, or make people comfortable with the fact that existing reporting may need to change. That is where leadership matters. If discrepancies are treated as blame, people will protect legacy numbers. If they are treated as a maturity issue, people are far more likely to surface the problem and work through it properly.
There is a difference between asking who got this wrong and asking how our reporting evolved in a way that created this gap. One creates defensiveness. The other creates accountability.
In the end, broad governance discussions are not enough. What usually helps is much simpler: clear owners, documented logic, metric-by-metric reconciliation, and visible change control when definitions shift. Because when reporting grows faster than accountability, every discrepancy becomes a rabbit hole. And usually, the rabbit hole is not the problem itself. It is the symptom of unmanaged metric evolution over time.
Governance sets the rules, ownership does the cleanup, reconciliation proves the truth, and change control protects the future. Without all four, the problem comes back.
Before asking for better governance, should we first be asking who actually owns the logic?
Data DataGovernance
No comments:
Post a Comment