Scenarios 01–05
1 — Tool-first lake
Situation: Executives buy a lake because competitors have one; no problem, value target, source plan, or success measure exists.
Primary problem: tool-first strategy failure.
Leading KA: Big Data and Data Science.
Supporting: Governance, Architecture, strategy/portfolio alignment.
Roles: sponsor, Platform Architect, Governance, Data Scientists.
Best response: define business need, measurable benefit, scope, sources, and strategy before committing the environment.
Weaker: “buy first to make progress” lets technology define requirements.
Changed fact: high-value measurable use case already exists; remaining question is capacity → environment/architecture selection becomes next.
Source: pp. 484–485, 496–497.
2 — Cheap but unreliable feed
Situation: Massive cheap social feed; sampling, continuity, and field meanings are unstable.
Primary: source veracity/semantic instability.
Supporting: Metadata, DQ, Governance.
Roles: Data Scientist, Curator, Metadata Specialist, Steward.
Best: evaluate origin, meaning, update behavior, consistency, reliability, bias; profile before selection/ingest.
Weaker: loading because it is large postpones the trust decision.
Changed: reliability/semantics/population proven → controlled ingest + Metadata/DQ becomes next.
Source: pp. 485–487.
3 — Lake without a catalog
Situation: thousands of raw files; analysts cannot tell origin, currency, or content.
Primary: swamp risk from missing inventory/Metadata.
Supporting: Metadata, Governance.
Roles: Metadata Specialist, Curator, Steward, Platform Architect.
Best: catalog source, structure, lineage, currency, definitions, relationships, intended use; govern future ingestion.
Weaker: more storage worsens discoverability.
Changed: Metadata exists but critical values are missing/inconsistent → DQ becomes primary corrective discipline.
Source: pp. 478–480, 499.
4 — Daily meets monthly
Situation: daily web behavior is joined to monthly customer value and treated as the same time grain.
Primary: temporal/granularity misalignment.
Supporting: Integration, Metadata, Master/Reference Data.
Roles: Data Scientist, SME, integration specialist.
Best: align timing/granularity and semantics; document transformations/assumptions.
Weaker: successful key join does not prove analytical alignment.
Changed: same time period/grain → investigate key/semantic/DQ alignment instead.
Source: pp. 487–488.
5 — Customer identity mismatch
Situation: clickstream cookie IDs, CRM customer IDs, service account IDs must become one customer view.
Primary: identity/context alignment.
Supporting: Master/Reference Data, Integration, Metadata.
Roles: Data Scientist, SME, MDM specialist, Steward.
Best: trusted common identifiers/entity resolution and MDM/RDM context.
Weaker: joining on names/convenient text creates false matches and bias.
Changed: all sources share a reliable enterprise customer key → identity resolution recedes; other alignment/DQ issues become primary.
Source: pp. 487–488.