Ontology is UnifyApps' master-data-management surface — where you define data models of your business entities, connect the sources that feed them, and let quality, matching, and survivorship rules consolidate everything into single, authoritative golden records.
Overview
Every business runs the same entity — a customer, a product, an account — through a dozen disconnected systems: a CRM, a billing platform, a support desk, a marketing tool. Each one has its own partial, sometimes conflicting version of that entity. Ontology exists to resolve that: you define what a "Customer" or "Product" looks like once, as a model, connect every source that has data about it, and let the platform reconcile the incoming records into one golden record per real-world entity.
Internally, Ontology is also referred to as MDM (master data management) — the same concept under its more common industry name. You'll find it in the sidebar under AI Ontology & Data › Ontology.
Key Concepts
| Term | What it is |
| Model | The top-level container for a complete data domain — holds entities, sources, and the relationships between them. |
| Entity | A business object definition inside a model — Customer, Product, Account — with its own fields, data quality rules, match rules, and survivorship rules. |
| Field | A single attribute on an entity (name, email, region). Fields can be primitive (String, Number, Date Time…), complex (Array, JSON, Lookup), or derived (Formula). |
| Relationship | A defined association between two entities — for example, a Customer places an Order. |
| Source | An external connection feeding data into an entity. Multiple sources can contribute to the same entity. |
| Golden Record | The unified, canonical record produced by merging every source's version of an entity through data quality, matching, and survivorship rules. |
How a Model Is Organized
Opening a model shows a left navigation grouped by the stage you're working in — the order roughly mirrors how you'd set one up end to end:
Data Modelling — Sources, Entities, Relationships: define the shape of your data.
Sync — Data Sync: bring records in, either through pipelines or automations.
Golden Records — Records (search the unified, consolidated records) and Quarantined Records, split into Potential Matches (suspected duplicates awaiting a decision) and Rejected Records (records that failed a data quality check).
Creating Your First Model
From the Ontology home screen, select New Model. You'll be asked for a Name (required) and a Description, plus the workspace or project it belongs to. From there:
Add one or more Entities, choosing whether each stores data in the JSON Store (document storage) or the Analytics Store (analytical storage) — this choice is fixed once made.
Define the entity's Fields, either manually or by importing them from a file, a connected source, or the template library.
Connect Sources and set up Data Sync to start bringing records in.
Add Data Quality, Match, and Survivorship rules so incoming records are validated, deduplicated, and merged the way you want.
Review the results in Records and Quarantined Records as data starts flowing in.
Note: Creating anything inside a model — a source, entity, relationship, field, match rule, or data quality rule — requires Edit access to the model itself. Without it, the corresponding "New…" or "Add…" action simply doesn't appear.
Use Cases
Customer 360 — Consolidate customer records scattered across a CRM, a billing system, and a support desk into one golden Customer record per person, so every team works from the same authoritative view.
Product Master Data — Model a Product entity fed by an ERP, an e-commerce platform, and a supplier feed, with match and survivorship rules resolving which source wins for each field when they disagree.
A Clean Data Layer for AI Agents — Give downstream automations, applications, and AI agents one consistent, deduplicated entity to query instead of reconciling raw source data themselves.
Notes
Keep the following in mind when working with Ontology:
An entity's storage choice (JSON Store vs. Analytics Store) is set at creation and can't be changed afterward — decide it deliberately.
A record that fails data quality or matching rules doesn't disappear — it lands in Quarantined Records for review, either as a Potential Match or a Rejected Record.
Sync brings data in through either pipelines or automations, not both at once per source — pick the mechanism that fits how that source's data becomes available.
Every "New…" / "Add…" action inside a model is gated on Edit access to the model, not just the object you're creating — a Viewer won't see the buttons at all.