Unify Data
Logo
Data Quality Rules

Data Quality Rules

Logo

4 mins read READ

Data Quality Rules let you validate, enrich, and clean up source records for a specific entity before they're matched and merged into golden records.

Overview

You add Data Quality Rules directly on an entity inside a model, from Ontology › [model] › Entities › [entity] › Data Quality Rules. Every rule you build here runs automatically on incoming source records during sync, before matching and merging happen — so problems get caught and handled while a record is still raw, not after it's already become part of a golden record.

This screen is different from the standalone Data Quality monitoring product elsewhere in the platform: that product scans and reports on data at rest, while the rules described here are entity-level checks you author yourself, wired into the ingestion pipeline of a single model. To create a rule you need Edit access to the model — without it, the New Data Quality Rule button simply won't appear.

ontology-data-quality-rules__p0.png
ontology-data-quality-rules__p0.png

Rule Types

Every rule you create belongs to one of three types, and for each type you can either build a custom rule from scratch or start from a ready-made template.

Rule TypeWhat it does
ValidationChecks incoming data for accuracy and correct formatting.
EnrichmentAdds or updates information on a record by pulling in additional data.
CleansingStandardizes data and removes inconsistencies, such as inconsistent casing or formatting.

ontology-data-quality-rules__s1-5566d8.png
ontology-data-quality-rules__s1-5566d8.png

Building a Rule

Click New Data Quality Rule and you'll move through a short stepper. The steps depend on the rule type you choose.

For Validation rules:

  1. Rule Type — choose Validation, Enrichment, or Cleansing, then decide whether to start from a custom rule or a template.

  2. Define Rule — configure the conditions and logic the rule checks. In Check Type, choose Standard to build simple conditions, or Automation to run an automation as the check. A rule uses one or the other.

  3. Actions Post Check — decide what happens to records that satisfy the rule's conditions (see below).

  4. Apply Filters — narrow which records the rule actually runs against, rather than applying it to every incoming record.

ontology-data-quality-rules__s2-e80e51.png
ontology-data-quality-rules__s2-e80e51.png

For Enrichment and Cleansing rules:

  1. Rule Type — choose the rule type, then decide whether to start from a custom rule or a template.

  2. Define Rule — configure the conditions and logic the rule runs. A rule can also run an Automation.

  3. Configure Output — map the automation's output parameters to entity fields so you control exactly where the enriched or cleansed data lands.

  4. Apply Filters — narrow which records the rule actually runs against.

Enrichment and Cleansing rules have no Actions Post Check step, so they don't offer the accept-with-warning or reject choice.

Actions Post Check: Accept with Warning or Reject

In the Actions Post Check step of a Validation rule, you set the action to be executed on records satisfying the rule, choosing one of two outcomes:

  • Accept records with warning — the record keeps moving through the pipeline, but it's flagged with a warning for reviewers to see later.

  • Reject record — the record is pulled out into Rejected Records for review instead of continuing on.

ontology-data-quality-rules__s3-169d5d.png
ontology-data-quality-rules__s3-169d5d.png

Whichever action you pick, you must write a warning or rejection message. Type {{ in the message field to insert dynamic data pulled from the record itself, so reviewers can immediately see why a given record was flagged rather than having to dig for the reason.

Note: if a rule's underlying Automation errors out while checking a record, the platform rejects that record automatically with its own generated message (something like "Automation execution failed: …") instead of using your authored text. If you see an unfamiliar rejection reason, it usually means the checking automation broke — not that the data itself is unusually bad.

Where Results Show Up

Rule outcomes surface in two places depending on the action taken. Records rejected by a rule land in the model's Quarantined Records › Rejected Records, where you can correct and re-submit them. Records that were merely warned show up as Data Quality Insights on the golden record itself — the Records screen displays an insights pill per record, and opening a record shows exactly which rules produced which insights during resolution.

Managing Existing Rules

The Data Quality Rules list shows every rule defined on the entity. From each row's menu you can edit a rule, clone it, or open a read-only details view. Deleting a rule asks for confirmation first and can't be undone, so double-check before you confirm.

Cloning a Rule

Use Clone from a rule's row menu to duplicate it quickly. This opens a dialog titled Clone [rule name] with:

  • A required Name field, prefilled as "[name] (copy)".

  • A Description field, prefilled from the source rule — edit either field before saving if you need to.

  • A Clone the associated automation checkbox, on by default. It appears only when the rule runs an Automation.

  • A Deploy the cloned automation checkbox, also on by default, which deploys the copied automation. It's hidden when you switch off Clone the associated automation.

For a rule that doesn't run an Automation, the dialog has only Name and Description.

ontology-data-quality-rules__s4-1c426a.png
ontology-data-quality-rules__s4-1c426a.png

Leave the automation checkbox on and a rule that runs an Automation as its check gets its own independent copy of that automation. Switch it off and the clone keeps pointing at the same automation as the original — meaning any later edit to that automation changes what both rules run. Click Clone to save the copy and jump straight to its edit page, or Cancel to close without creating anything.

The clone is always created disabled, regardless of whether the source rule was on or off, so it won't run until you switch it on from the list. Cloning requires the same Edit access as the rest of the Data Quality Rules screen.

Notes

  • Rules run during sync, before matching and merging — they act on source records, not on already-formed golden records.

  • Without Edit access to the model, you won't see the New Data Quality Rule button or the Clone action.

  • An unexpected, platform-generated rejection message almost always points to a failing Automation, not bad source data.

  • Cloning with the automation checkbox off links two rules to one automation — changing that automation later affects both rules at once.

  • Deleting a rule is permanent; there's no undo after you confirm.

FAQs

Can one rule use both Enrichment logic and a custom condition?

No. Each rule has one Rule Type (Validation, Enrichment, or Cleansing), and the type determines its steps. Enrichment and Cleansing rules get a Configure Output step to map automation outputs to fields, and have no Actions Post Check step. In a Validation rule, the Check Type is either Standard (conditions you build) or Automation, not both, and the rule then gets an Actions Post Check step. To check records in different ways, create separate rules.

Do rules apply to every record, or can I scope them?

Use the Apply Filters step to restrict a rule to a subset of records instead of running it against everything coming into the entity.

What happens to a cloned rule immediately after I save it?

It's created disabled and you're taken straight to its edit page, so you can review or adjust it before turning it on from the rules list.

Where do I find records a rule rejected?

They land in Quarantined Records › Rejected Records on the model, where you can correct and re-submit them.