Find unused calculated fields in a workbook.

What is the direct answer to this workbook question?

To find unused calculated fields in a .twb or .twbx workbook, build a dependency graph linking every calculation's internal ID ([Calculation_...]) to the worksheet shelves, filters, parameters, and nested formulas that reference it. Any field with zero downstream references is an orphan. Drop your .twbx into the xmetric inspector to check fields locally.

Why check unused calculated fields before migration?

Production Tableau workbooks accumulate dead weight over years of edits: Sales (copy), Temp_Test_LOD, Old Margin v2, and broken experiments that no worksheet displays.

When your team migrates a .twbx workbook to SQL or dbt, failing to prune unused calculated fields first means you waste hours translating dead formulas that never appear on any dashboard. Worse, materialized extract refreshes can still evaluate unused row-level calculations if they aren't cleaned up.

Why does text search miss nested dependencies?

You cannot find unused fields just by checking if a field's display name appears on a worksheet shelf:

1. Internal IDs vs Captions: Inside the .twb XML, a field named Net Revenue in the UI is referenced as [Calculation_839201928374] inside other formulas and shelf tags.
2. Transitive (multi-hop) dependencies: Calculation A might not sit on any worksheet shelf directly, but Calculation B (which is on the Columns shelf of Sheet 3) calls Calculation A inside an IIF() statement.
3. Hidden shelf usages: A calculated field can be unused on Rows and Columns while still driving a Context Filter, a Sort order, a Tooltip, a Reference Line, or a Color mark.

How can a Python script find unused calculated fields?

Run this script directly or hand it to Claude Code or Cursor to list every true orphan calculation in your workbook:

import re, sys, zipfile
import xml.etree.ElementTree as ET

def find_unused_calcs(path):
    if path.endswith(".twbx"):
        with zipfile.ZipFile(path) as zf:
            twb = [n for n in zf.namelist() if n.endswith(".twb")][0]
            xml_text = zf.read(twb).decode("utf-8", errors="ignore")
    else:
        xml_text = open(path, encoding="utf-8", errors="ignore").read()

    root = ET.fromstring(xml_text)

    # 1. Collect all calculated fields: internal_name -> caption & formula
    calcs = {}
    for col in root.findall(".//datasource/column"):
        calc = col.find("calculation")
        name = col.attrib.get("name")
        if calc is not None and name and "formula" in calc.attrib:
            calcs[name] = {
                "caption": col.attrib.get("caption", name),
                "formula": calc.attrib["formula"],
            }

    # 2. Collect all text in worksheets, dashboards, and actions
    ws_text = " ".join(
        ET.tostring(el, encoding="unicode")
        for el in root.findall(".//worksheets") + root.findall(".//dashboards") + root.findall(".//actions")
    )

    # 3. Mark directly used calcs, then walk transitive formula dependencies
    used = {cid for cid in calcs if cid in ws_text}
    changed = True
    while changed:
        changed = False
        for cid in list(used):
            formula = calcs[cid]["formula"]
            for other_id in calcs:
                if other_id not in used and other_id in formula:
                    used.add(other_id)
                    changed = True

    unused = [
        (cid, info["caption"], info["formula"])
        for cid, info in calcs.items()
        if cid not in used
    ]
    return unused

for cid, caption, formula in find_unused_calcs(sys.argv[1]):
    print(f"UNUSED: {caption} ({cid}) -> {formula}")

How can I inspect and verify workbook fields in my browser?

Before deleting a list of orphan fields or skipping them in a dbt migration, drop your .twbx or .twb file into the xmetric workbook inspector in your browser:

• See every worksheet, data source, and calculation in one place. The inspector traces what each calculation depends on, shows fields nothing uses, and flags calculations that refer to a missing field. No Tableau Desktop install needed (Tableau offers a free local edition for one person without publishing or sharing, or Tableau Creator at $75 per user/month, billed annually as of October 2026).
• Save packaged data and test formulas locally. Save the data files packaged inside a .twbx, or run formulas in the formula playground in your browser. Nothing is uploaded.
• Agent-ready inspection. The public xmetric MCP server is live at https://getxmetric.com/mcp with no authentication or API key. Anonymous initialize, tool discovery and widget-resource checks passed on 2026-10-29. ChatGPT and Claude host rendering and tool selection remain unverified; their directory listings remain coming. Copy the browser inspector output for an unused-field audit; the public MCP catalog does not provide an unused-field audit tool.

What is live, and what remains planned?

The public xmetric MCP server is live at https://getxmetric.com/mcp with no authentication or API key. Anonymous initialize, tool discovery and widget-resource checks passed on 2026-10-29. ChatGPT and Claude host rendering and tool selection remain unverified; their directory listings remain coming. See the dated public protocol receipt and setup and evidence limits.

The public server is available; host operation and directory approval are separate checks.

The file board parses supported .pbix, .twbx/.twb, CSV, Excel and Parquet files in your browser. Choosing a file shares supported schema, source names and calculation definitions with xmetric and the chat host; requested result cells, filters and SQL or DAX evidence are shared when answering. The original file is not uploaded by this flow. Small aggregates and labels can still be sensitive. Privacy details.

What data does this worked example use?

The initial view uses authored synthetic data, not customer data or a captured agent session.

The example record lists its file, calculation or SQL and expected values. The live browser result is computed by xmetric’s existing engine or workbook inspector. Samples are excluded from own-data activation reporting.