Marmot Column Tags

[tdari]

Hey @charlie-haley 👋 ,

I started initial implementation of column tags based on our conversation here.

I’d like to get your feedback and any considerations before moving forward. We can discuss and shape the initial functionality together.

I’ll likely have more questions later, but to start:

• Should or could we consolidate asset tags and column tags? (I’d prefer that if possible, but I need to investigate and confirm first.)

---

Feature Summary

Column Tags are a structured tagging system that allows users to classify and label individual columns in database schemas. Instead of tagging entire assets, column tags provide column-level metadata that describes the sensitivity, quality, compliance status, and other attributes of specific data fields.

Tag Categories

A tag category is a dimension or aspect used for classification. Categories represent different types of information you want to track about columns. Each tag category defines a context for a tag value. For instance, an "internal" tag might have different meanings in different categories.

Examples:

• Sensitivity: Describes how sensitive the data is
• PII: Identifies what kind of personally identifiable information it is
• Quality: Indicates data quality level
• Compliance: References regulatory requirements

Tag Values

A tag value is a specific option within a tag category. Values are the labels applied to columns.

Examples for Sensitivity category:

• Confidential: Highly sensitive business data
• Internal: Internal use only
• Public: Safe to share externally

Examples for PII category:

• Email Address
• Phone Number
• Social Security Number
• Home Address

---

How Column Tags Work

Tag Category and Value Management

• We will add a new section in the Governance tab called "Tags"
• Users will be able to manage their tag categories and values.
• All tags, including asset tags, will be managed from here.

Why:

• This reduces duplicate tags such as "PII" and "pii" or "Production" and "Prod", and provides a single source of truth with clear tag descriptions.

Here are some draft UI designs for Tag Governance (not final):

---

---

The Tagging Process

1. A user opens a schema in Marmot.
2. For each column, they assign relevant values from available categories. (There will be some pre-populated tag categories and values for common use cases e.g., PII)
3. Example: customer_email column gets tagged with:
   • PII: Email

Here are some draft UI designs for Column Tagging flow (not final):

Storage

• Tag Values and Tag Categories will be stored in separate Postgres tables.

Discovery & Filtering

Users can find assets by their column tags using the filter panel and search bar on the Discover page.

---

Example Use Cases

• Identify which assets have PII columns
• Automatically redact PII values in table previews.
• Enrich asset tags using asset rules (e.g., if a column contains "phone number," assign a "PII: Phone Number" tag).

[charlie-haley]

Sounds great, keeping tags on both columns and assets would be good. Initially, we can keep simple tags for columns without further classifying e.g with PII and having Name/Emaill/Address, I think for now the single tag will be easier.
If we could make the tag field free-text (similar to GitHub) but, reusable once created, I think that would be a nice addition

Some plugins also support classifying columns (such as DBT), so maybe we can factor that in, doesn't have to be as part of the initial implementation.

[tdari]

I agree. So for the initial phase:

• No tag categories, just tag values are sufficient.
• Tag values should be reusable.
• Asset and column tags should share the same tag entities.
• If possible, tag values should be synced from asset metadata tags (e.g., dbt).

edit: Removed my question. You've already answered it. thank you.