Wiki Navigation

Explore and Understand a Model

This workflow is the fastest way to get oriented without running large queries or making changes.

It’s designed for analysts, developers, and consultants who need a mental model of:

  • what the model contains,
  • how it’s structured,
  • where key business logic lives,
  • what might be risky or inconsistent.

The “explore” workflow (recommended sequence)

01
Inventory
High-level summary
02
Search
Business terms
03
Deep Dive
One table at a time
04
Scan
Logic & security

  1. High-level inventory

    “Summarize the model: main tables, relationships, measures, calculation groups, roles, and any obvious red flags.”

    Ask for:

    • likely fact tables,
    • likely date tables,
    • relationship patterns (inactive, bi-directional),
    • duplicated time-intelligence patterns,
    • suspicious naming/hidden-field patterns.

  2. Focused search (business term → model objects)

    “Search for anything related to ‘customer churn’ in names and descriptions first. Then, if needed, search expressions.”

  3. Table deep dive (one table at a time)

    “For table ‘Sales’: show key columns, data types, hidden fields, and relationships.”

    If you’re doing performance work:

    “Flag high-cardinality columns and suspicious relationship directions.”

  4. Semantic logic scan (measures/calc groups)

    “List calculation groups and explain what they do.”“List top measures (by naming/folder conventions) and highlight missing descriptions.”

  5. Security scan (if relevant)

    “List roles and summarize filters/permissions at a high level.”

Tips that keep results useful

Browse-only playbook (when queries/previews are disabled)

Common “consulting” prompts

“Generate a model review checklist and then assess this model against it (naming, relationships, date table, calc groups, security, performance signals).” “Create a glossary of key measures: name → description → how to use → common pitfalls.” “Identify unused or suspicious objects (measures never referenced, hidden tables with many columns, etc.).”

Common deliverables (what the assistant can produce)

  • Model overview (tables, relationships, key measures)
  • Measures glossary (name → description → how to use)
  • “Where is X implemented?” inventory (search results grouped by type)
  • Risk list (inactive relationships, missing date table, duplicate TI logic)

Troubleshooting

You may be in browse-only mode. Ask the assistant to switch to metadata-only exploration and propose follow-up steps.

Ask to restrict types: “measures only”, “tables + columns only”, or “one table only”.

See also