The Family Tree in the Filing Cabinet

Opening Scene

A registry doesn’t stop at giving each citizen their own file. It also tracks that this citizen is the child of that citizen, that this household consists of these four people at this one address, and that a business registered downtown is actually a subsidiary of a parent company registered in another city entirely. Get any of those relationships wrong and the consequences ripple outward — a tax notice sent to the wrong household, an inheritance record filed against the wrong parent, a regulatory filing that misses an entire subsidiary because the parent-child link was never recorded. The individual file cards were never the whole story. The relationships between them were always part of what the registry existed to maintain.

In Plain English

Master data hierarchies and relationships capture how master entities connect to one another — not just what each entity is, but how it fits into a larger structure. Common patterns include parent-child hierarchies (a corporate subsidiary rolling up to a parent company, a product variant rolling up to a base product), household or grouping relationships (multiple customer entities linked as one household or account family), and cross-entity relationships (a specific customer linked to the specific vendor who fulfills their orders). Getting entity resolution right, as covered in Articles 3 and 5, only solves half the problem — a golden record for each individual entity is necessary but not sufficient if the relationships between those entities are missing, stale, or wrong.

The Old Way

Hierarchy and relationship data has always been harder to maintain well than individual entity data, for a few consistent reasons:

Hierarchies change more often than the entities inside them — a subsidiary gets acquired, sold, or restructured; a household member moves out; a product gets reclassified under a different parent line — and every one of these changes requires updating the relationship, not just the entities on either end of it.
Different systems capture different, often incompatible, views of the same hierarchy — a sales system might roll a customer up to an account team’s territory, while a billing system rolls the same customer up to a different corporate parent for invoicing purposes, and both views can be simultaneously legitimate for their own purpose.
Circular or inconsistent hierarchies are a persistent risk — a subsidiary accidentally listed as its own parent, or a household loop where two customer records each claim to be the other’s dependent, tends to surface only when a hierarchy-dependent report or process breaks in some confusing way.

Maintaining hierarchy data well has always required treating the relationships themselves as first-class data, with their own validation and stewardship, rather than as an afterthought bolted onto individual entity records.

What’s Changing (and Why AI Is the Reason)

AI-assisted relationship inference can propose likely hierarchical links — such as detecting that two customer accounts share enough signals (shared billing address, shared contacts, shared payment method) to plausibly belong to the same corporate family — even when no system has ever explicitly recorded that link. This surfaces genuine structural relationships that pure entity-level matching, focused on single entities in isolation, would never catch.
AI models can reconcile competing hierarchy views from different systems more gracefully than a rigid rules engine, weighing which view is more appropriate for a given downstream use case rather than forcing a single, universal hierarchy onto every consumer. This respects the reality that a sales-territory rollup and a billing-parent rollup can both be legitimate at once, rather than treating one as simply wrong.
AI agents that reason over master data increasingly depend on relationship context to answer questions correctly — an agent asked “what’s this corporate family’s total spend?” needs a reliable hierarchy to even define the boundary of “this corporate family” before it can begin the calculation, making hierarchy quality a direct precondition for a whole class of AI-assisted analysis.

The Metaphor, Fully Extended

Registry Element	Master Data Management Concept
A parent-child link between a citizen and their child’s own file	A hierarchical relationship between two master entity records
A household file grouping four individual citizen files at one address	A grouping relationship linking multiple entities into one logical unit
A downtown business’s file noting it as a subsidiary of an out-of-city parent	A corporate hierarchy connecting related business entities
Two offices each maintaining a different, valid view of the same family’s structure	Competing but legitimate hierarchy views held by different source systems
A registry assistant noticing two business files share enough signals to likely belong to one family, unrecorded anywhere	AI-assisted relationship inference, surfacing undocumented structural links

For Beginners: What to Actually Do

When you encounter a master data hierarchy, ask who maintains it and how often it’s reviewed — a hierarchy nobody actively maintains degrades faster than most people expect.
Practice distinguishing entity-level data quality from relationship-level data quality — a perfectly accurate golden record can still sit inside a completely wrong hierarchy.
Get comfortable with the idea that two systems can both be “right” about a hierarchy while disagreeing with each other, because they’re answering different questions.
Watch for circular or self-referencing relationships as a specific, recognizable failure mode worth checking for directly, not just a generic data quality concern.

For Practitioners and Leaders: The Deeper Layer

Treat relationship and hierarchy data as a first-class governance concern with its own stewardship ownership, not an implicit byproduct of individual entity records.
Maintain multiple hierarchy views deliberately where genuinely needed by different business functions, rather than forcing a single hierarchy to serve every use case poorly.
Use AI-assisted relationship inference to surface undocumented structural links as candidates for steward review, particularly for corporate family and household relationships that tend to go unrecorded.
Recognize that hierarchy quality directly gates the reliability of any AI agent or analysis that reasons about entities in aggregate, not just entities individually.

Quick Recap

Master data hierarchies and relationships capture how entities connect to one another — parent-child, household, and cross-entity links — beyond what any individual golden record describes alone.
Hierarchies change frequently, differ legitimately across systems, and are prone to circular or inconsistent structures if not actively maintained.
AI-assisted relationship inference can surface undocumented structural links, and AI models can reconcile competing hierarchy views more gracefully than rigid, one-size-fits-all rules.
Hierarchy quality is a direct precondition for reliable AI-assisted analysis that reasons about entities in aggregate, not just individually.

Where This Fits in the Series

Articles 1 through 5 focused on individual entities. This article extended the picture to the relationships between them. Article 7 turns to the specific domain of customer master data — the most common starting point for most MDM programs.