Lesson 2 — Inmon / Corporate Information Factory
The four classic warehouse properties
The Inmon definition is a high-probability exam target:
Subject-oriented · Integrated · Time-variant · Non-volatile
Subject-oriented
Operational systems are often organized around applications/functions: billing, claims, order entry. A warehouse reorganizes information around enduring business subjects such as Customer, Product, Account, or Order so analysis can cross application boundaries.
Integrated
Integration is not copying. Different sources may encode the same concept with different keys, names, code sets, units, or definitions. Warehousing reconciles those differences into a common analytical representation.
Example: US, USA, and 840 may all represent the United States. Copying all three unchanged preserves inconsistency. Integration applies a governed mapping so the analytical meaning is consistent.
Time-variant
The warehouse preserves states over time. If a regulator asks what was reported for Q2 last year, the design should be able to reproduce that point-in-time result rather than answer only with today’s values.
Non-volatile
Historical warehouse data is stable compared with operational data. Operational records may be repeatedly updated in place; warehouse history generally preserves prior states and adds new states rather than destroying evidence.
Non-volatile does not mean “no more data can ever be loaded” or that every storage engine is literally read-only.
Corporate Information Factory (CIF)
The current source emphasizes a broad architecture containing:
Applications/Sources → Staging → Integration/Transformation → Enterprise DW → Data Marts → Analysis
An ODS can sit near the integration path to support current/near-current operational needs. Reference/Master/external data provide consistent shared context.
The defining Inmon clue is the central enterprise warehouse as the principal integration point, traditionally normalized, with downstream marts/presentation structures able to consume it.
ODS vs historical DW
This distinction is frequently tested.
ODS - integrated; - current or near-current; - lower latency; - shorter history; - more volatile; - useful for tactical/operational analysis.
Data Warehouse - integrated; - durable historical depth; - more stable/non-volatile; - trend, compliance, strategic and reproducible analysis.
A 30-day current service view refreshed every few minutes points toward an ODS. Five years of integrated trends points toward the DW.
Operational Data Mart
An Operational Data Mart (OpDM) inherits the current/near-term character of the ODS and supports tactical decisions. A conventional historical mart usually presents a subject/process subset from the historical warehouse or related analytical layer.
Source defects
Warehouse transformation should not become a permanent hiding place for bad source data when the source can be corrected. The durable pattern is: control the warehouse load, document what happened, and feed defects upstream for remediation where practical.
Stop and check
Why is an ODS not simply a smaller DW? Because the deciding attributes are latency, volatility, time horizon, and operational purpose—not size.
Scenario: A company wants one normalized enterprise customer/order integration layer that feeds Finance and Sales marts. Answer: Inmon/CIF is the stronger clue.
Source anchor: pp. 365–368.