Someone asks whether a database holds personal data. The question is about the database; the answer lives on its columns, three relations down, scattered across hundreds of assets. The catalog already knows the answer. Now you see it where it matters most.
What's new: Derived attributes
Derived attributes are a new kind of attribute in Collibra, now in private preview. Instead of someone typing a value in, Collibra calculates it. It follows a path of relations from the asset you're looking at, collects one attribute from the assets at the far end and applies a single rule to what it finds: the list of unique values, show “true” if any value is true, a count, a sum, an average, the lowest or highest value.
How derived attributes help
Governance questions get asked at the level where decisions happen: the AI use case, the data product, the report. The facts that answer them are usually recorded somewhere else: on the column, the table, the source system. Collibra holds the relations between those assets, so the information is there. Until now, getting it to the level where someone needed it meant sifting through a hundred assets or creating custom reports.
Problems it solves
- The answer sits on another asset: sensitivity flags live on columns, but the question is about the database or the data product that uses them.
- Checking by hand doesn't scale: confirming a database-level answer means opening schema, table and column pages one by one, and hundreds of columns turn a simple question into an afternoon.
- Copied values go stale: teams copy the answer onto the parent asset, then someone has to remember to update it every time a column is tagged or a table moves.
- Propagation workflows are a second job: scripts and workflows that push values up the hierarchy take effort to build, break when relations change and add extra values to the graph.
- People stop asking: when every cross-asset answer costs that much, teams settle for the attributes they can see directly and leave the rest of the graph unused.
How derived attributes work
Take the question from the top: does this database contain PII?
Today, you open the database in Collibra, click into a schema, open a table, then scroll its columns looking for the Personally Identifiable Information flag. Then you do it again for the next table. Or, more likely, someone on your team added a "Contains PII" field to the database months ago, and you're hoping they kept it current.
With derived attributes, an administrator sets this up once. In the operating model settings, they add a derived attribute type called Contains PII to the Database asset type and give it three things:
- Path: Database → Schema → Table → Column
- Rule: true if any column is true
In plain English, that says: start at this database, walk down through every schema and table to every column and check each column's PII flag. If even one says true, the database contains PII. Now you open any database page and Contains PII is right there with the other attributes, already computed.
With derived attributes information is shown where it’s needed.
This works because the relations are already governed. The path from database to column exists in the Collibra catalog.
Why you should be excited
- Operating model administrators: define a business question once as a derived attribute type and retire the workflows and scripts that copy values between assets.
- Data stewards: open a database and see whether it contains PII without opening a single column page. Time that went into clicking through the hierarchy goes back into curating the columns themselves.
- Data product owners: see what a data product is built from, such as the source system types behind it, on the product's own page instead of tracing its columns back to their databases.
- Privacy and security teams: get the database-level view of where personal data sits, computed from the column-level flags stewards already maintain.
Use cases
- AI trained on non-sensitive data: Contains PII derived attribute on an AI use case returns “true” if any column's Personally Identifiable Information flag is true. When a steward tags a new column, the AI use case updates after a short delay with no one touching the database.
- Data source types behind a data product: a derived attribute on Data Product follows Data Product → Column → Table → Schema → Database and returns the distinct values of the Data Source Type attribute held at database level. A consumer opening the data product sees which kinds of sources feed it without walking the relations back to each database.
- Column count on a database: a derived attribute on Database counts the columns beneath it and shows the total on the asset page. An owner sizing up a database gets a number instead of a scroll.
Key takeaways about derived attributes
Derived attributes move cross-asset answers onto the asset where the question gets asked. Values are calculated rather than copied, so there's no duplicate to keep in sync and no custom workflow to maintain. The capability is in private preview and soon it will move to public preview.
- Ask the question where it lives: the answer appears on the database, schema or data product page, not a hundred clicks down.
- Define once, compute everywhere: one definition on an asset type applies to every asset of that type.
- No manual upkeep: derived values come from the source attributes stewards already maintain, so there's nothing extra to reconcile.
- Maxime JeanmartMaxime JeanmartProduct Manager, Data GovernanceCollibra
Maxime Jeanmart
Maxime Jeanmart
Product Manager, Data Governance
Collibra
In this post:
- What's new: Derived attributes
- How derived attributes help
- How derived attributes work
- Why you should be excited
- Key takeaways about derived attributes
Keep up with the latest from Collibra
I would like to get updates about the latest Collibra content, events and more.
Business e-mail address
There has been an error, please try again
Please enter a valid business e-mail address
By submitting this form, I acknowledge that I may be contacted directly about my interest in Collibra's products and services. Please read Collibra's Privacy Policy.
Thanks for signing up
You'll begin receiving educational materials and invitations to network with our community soon.











