Lesson 10 — Governance, Acceptance & Metrics
Governance should control risk without becoming obstruction
DW/BI governance establishes decision rights and controls for subjects such as: - authoritative sources and integration boundaries; - staging/refinement/production separation; - Data Quality expectations and remediation; - Metadata and lineage; - security, privacy, access and retention; - self-service/discovery and publication; - release/configuration controls; - service expectations; - exceptions and escalation.
The exam trap is treating governance as either “no controls” or “approve every query.” The intended model is business-driven, risk-proportionate control.
Readiness can stop a program
Current source guidance treats missing business commitment as serious. If the business will not supply sponsorship/SMEs for definitions, rules, validation, and acceptance, this is not merely an IT inconvenience. The initiative may need to stop until commitment exists.
Other readiness dimensions include source/data quality, architecture/platform capacity, skills/staffing, security/privacy/legal constraints, Metadata/lineage, production support, performance/availability, and service expectations.
Business acceptance
Technical loading success is not enough. Trust requires:
Understandable data + verifiable Data Quality + demonstrable lineage + business validation/sign-off.
UAT should compare BI results against source-system evidence over the initial load and subsequent update cycles—not just confirm a dashboard opens.
Service Level Agreements
SLAs define agreed service behavior appropriate to the environment: availability, load/delivery windows, freshness/latency, response/support, recovery, retention, or other producer-consumer expectations.
An ODS and historical warehouse may legitimately have different SLAs.
Reporting strategy
A reporting strategy is broader than a BI tool inventory. It considers: - security/access; - user groups and tool fit; - report/analysis type; - frequency and distribution; - storage/retention; - visualization; - timeliness vs performance trade-offs; - support and publication expectations.
Center of Excellence
A BI/DW Center of Excellence can provide training, reusable patterns/templates, source guidance, tool expertise, standards/governance communication, support practices, and coordination across teams.
Metric families: ask the management question
Usage / adoption
Are people actually using it? Connected users, concurrent users, query/report activity, frequency, trends. License/registration counts show entitlement/capacity, not behavior.
Subject-area coverage
How broad is the intended analytical scope? Which domains/sources/departments are represented and actually consumed?
Response / performance / load support
Does the service meet expected windows? Load duration/success, refresh timeliness, query response, extract completion, availability.
Satisfaction / trust
Do users perceive the product as useful, reliable, understandable, and supported?
No one metric proves total value. “Five thousand registered users” says little without actual activity, coverage, service performance, and user value.
Final mental model
Business need → history/grain/latency → source + DQ reality → architecture → integrate/store/present → BI delivery → Metadata/lineage/Governance/Security → releases/operations → usage/coverage/performance/satisfaction → improvement.
Source anchor: pp. 387–393.