Every object in Data Quality has its own profile page, where you can inspect field-level statistics, preview the underlying records, and attach evaluation rules that check the data going forward. This page walks through the Fields tab, the Preview Data view, and the Rules tab, including how AI and system-generated suggestions work, how to map a rule's inputs, and how Copilot can build and attach rules for you conversationally.
Overview
Open any object from your Data Quality suite to land on its profile. Two tabs sit at the top: Fields, which shows profiling results for every field in the object, and Rules, where you review, suggest, and attach the evaluation rules that will score the object's records. Together they answer two different questions — what does this data actually look like, and what rules should check it — so it's worth treating them as separate steps rather than rushing straight to rule creation.
Review field statistics
The Fields tab opens on a profiling report. At the top you'll see headline counts for the object: total fields analyzed, records processed, and duplicate records found. Below that is a table listing every field, its linked business terms, its null rate, and its distinct versus non-distinct value counts, alongside a distribution chart so you can spot skew or outliers at a glance.
Expand any field row to see statistics tailored to its data type. Numeric fields show mean, minimum, and maximum values; text fields show average, minimum, and maximum length along with an empty string count; boolean fields show the true and false percentage split. Use these numbers to sanity-check a field before you build a rule around it — a field with an unexpectedly high null rate or an odd distribution is usually worth investigating first.
Each field row also has a control for linking business terms. Profiling automatically detects likely terms based on your Business Ontology's detection rules, and you can add or correct the linked terms manually from the same row. This matters beyond labeling: a term's evaluation rules are meant to follow every field it's linked to, so accurate term links help the right rules find the right fields later.
Preview the underlying data
Statistics tell you the shape of the data, but sometimes you need to see the records themselves. Use Preview Data on the object to view a sample of up to the first 100 records. Treat this as a spot check rather than a query tool — it's there to confirm fields are mapped correctly and that values look reasonable before you start trusting rule results built on top of them.
Review AI and system suggested rules
Switch to the Rules tab to work with evaluation rules, and you'll find two suggestion streams shown side by side so you can compare them before deciding what to add.
AI Suggested Rules are generated from your data plus any context you supply through Add Instructions or Setup Instructions. Each suggestion carries a confidence percentage so you can gauge how strongly the AI backs it. If you haven't added any instructions yet, you'll see a prompt to add rule knowledge so the AI has enough context to generate relevant suggestions; while it's working, it shows that it's actively crafting rules for you.
System Suggested Rules are derived directly from the profiling results on the Fields tab — things like null-rate or distinctness checks that the platform can infer without extra input. If none apply yet, you're prompted to add your own evaluation rules instead.
Select the suggestions you want, use Add {n} Rules to bring them into the object, then review the field mappings for the rules you selected before they take effect.
Map rule inputs and scope with filters
Attaching any evaluation rule — whether it came from a suggestion or you added it manually — opens the input mapping step. Here you select which fields on the object should serve as inputs for each of the rule's attributes and parameters; a rule can also be scoped to a fixed value instead of a field where that makes sense.
Optionally, apply rule filters to set conditions that determine when the rule should run at all. This is useful when a rule only makes sense for a subset of records — for example, only where a status field equals a certain value. Records that don't match the filter aren't scored pass or fail; they come back as skipped in your scan results, which keeps them from skewing your pass rate.
Use Copilot to create and map rules
If you'd rather describe what you need than configure it field by field, open Ask Copilot from the object. It's a chat interface that can create and map evaluation rules conversationally, with starter prompts such as asking for help creating an evaluation rule, detecting duplicate records, or setting data freshness rules. Copilot proposes both the rule logic and the field mappings, and you review and apply its proposals the same way you would a manual suggestion — nothing is attached without your confirmation.
Notes
Preview Data only ever shows up to the first 100 records, so it's not a substitute for running an actual scan.
A field's linked business terms can pull in that term's evaluation rules automatically, so keep term links accurate as you profile new objects.
Filtered-out records show up as skipped rather than failed, so a low pass rate is worth checking against your rule filters before you assume there's a data problem.
AI Suggested Rules improve as you add more context through Add Instructions or Setup Instructions — an empty knowledge base means fewer, less confident suggestions.