The Data Warehouse That Nobody Trusts
A retail bank has invested £4m in a new enterprise data warehouse (EDW) to support regulatory reporting and management dashboards. Six months after go-live, the finance team refuses to use the EDW because the customer balance figures differ from the core banking system by up to 3%. The EDW team claims the core system is wrong. An investigation reveals that the EDW's ETL process aggregates balances at month-end, but the core system updates balances in real-time. The EDW also includes legacy accounts that were closed but not purged. The data governance council, which meets quarterly, has not defined data ownership or data quality standards. The bank's regulator has threatened to impose a Section 166 skilled person review if the next regulatory report (due in three months) is inaccurate. The EDW project manager wants to hire more developers to rebuild the ETL. The CDO argues that the problem is not technical but organisational.
1. Analyse the root causes of the data quality issues and propose a remediation plan that addresses both technical and organisational dimensions. Prioritise actions for the next three months.
Read the worked answer
Root causes: (1) Technical: ETL timing mismatch (month-end vs real-time) causes reconciliation differences; legacy accounts not purged cause overstatement. (2) Organisational: lack of data ownership and quality standards; no data governance council empowerment; unclear accountability between EDW and core system teams. The CDO is correct: the issue is organisational as much as technical. Remediation plan: Immediate (next 3 months): (a) Establish a data quality task force with representatives from finance, IT, and data management, reporting to the CDO; (b) Define critical data elements (CDEs) for regulatory reporting, assign data owners, and document data quality rules; (c) Implement a reconciliation process between EDW and core system, with automated exception reporting; (d) Purge or flag legacy accounts in the EDW, with a clear business rule; (e) For the upcoming regulatory report, use a manually reconciled extract from the core system as a temporary control, with sign-off by the CFO. Medium term: (a) Define data governance policies, including data ownership and stewardship; (b) Align ETL timing with business needs, possibly moving to intra-day updates; (c) Invest in data quality monitoring tools. The project manager's proposal to hire more developers is insufficient without addressing governance. The three-month priority is to ensure the regulatory report is accurate, which may require manual workarounds, while putting in place the governance foundation.
2. Explain how the bank should structure data ownership and stewardship to prevent recurrence, and how this aligns with the concept of data as an asset.
Read the worked answer
Data ownership should be assigned at the business unit level for each critical data element. For example, the finance department should own customer balance data, with a named data owner accountable for data quality. Data stewards should be appointed within business units to implement data quality rules, monitor compliance, and resolve issues. The data governance council should be elevated to a steering committee with decision-making authority, meeting monthly initially. This structure aligns with treating data as an asset: data has value, risk, and lifecycle. Ownership ensures accountability for data quality, similar to financial assets. It also supports regulatory compliance (e.g., BCBS 239 for risk data aggregation). The bank should include data quality metrics in performance objectives for data owners. This organisational change, combined with technical fixes, will prevent recurrence by embedding responsibility for data quality into business processes rather than treating it as an IT problem.