Department Anomaly Scan — GL Miscoding Cleanup

Last updated: April 17, 2026

Scans your workspace for GL-to-department coding anomalies — expenses hitting the wrong cost center, vendors consistently miscoded, accounts landing in unexpected departments. Surfaces patterns and generates a NetSuite reclass CSV journal entry to fix them.

Before You Begin

  • Numeric MCP connected.

  • Numeric Toolkit installed.

  • An income statement report with a department pivot (or access to build one ad-hoc).

  • Access to transaction-line drill-down.

Default anomaly rules are tuned to a hardware/software company (80xxxx = R&D, 70xxxx = S&M, 670xxx = G&A). Claude adapts to your actual department structure based on the report data.

How It Works

  1. Ask Claude to scan. Example prompts:

    • "Scan for department anomalies."

    • "Find GL miscodings."

    • "Anything miscoded this period?"

    • "Run a department audit."

  2. Claude lists available IS reports; you pick one (or choose ad-hoc).

  3. Claude validates the report has department data; if not, asks for a different report.

  4. Claude applies anomaly rules — GL ranges vs expected departments, new large items ($0 last month → >$10K this month), extreme variance (>500% MoM), round-number manual JEs.

  5. Claude drills into transaction lines on flagged rows to confirm root cause (card default, accrual template, manual JE).

  6. Claude generates a balanced NetSuite CSV with debit-correct-dept, credit-wrong-dept entries.

What You Get

  1. Anomaly table in chat — flagged account × department combinations with drill-down notes

  2. NetSuite CSV file: External IDs RECLASS-YYYY-MM-NNN, one JE per anomaly

  3. Verification: programmatic DR = CR balance check per JE with PASS/FAIL status

  4. Entries flagged [REVIEW W/ MGMT] where intent is ambiguous (e.g. PR spend under CEO)

Default anomaly rules:

GL Account

Expected Departments

Anomaly If Found In

80xxxx–81xxxx (Hardware Dev, Manufacturing, Lab)

R&D

S&M, G&A, COGS

70xxxx (Brand, PR, Performance Media)

Sales & Marketing

R&D, G&A

54xxxx (COGS, Reliability Testing)

COGS / R&D

S&M, G&A

650130 (Legal Fees)

Legal

Any non-Legal

651100 (HR Expense)

Talent / HR

Any non-Talent

670xxx (Lease, Rent, Utilities)

G&A

R&D, S&M

Troubleshooting

Issue

What to Check

Report doesn't have dept data

Won't proceed. Build ad-hoc with pivots: ["Department", "account"].

Rules don't match your company

Default rules are a starting point. Claude adapts based on actual department names / GL ranges in your workspace.

All anomalies from one vendor

Often the vendor's default card department is wrong in NetSuite — fix at source before running reclass.

FAQ

  • Does this skill work on any company? Yes — the skill reads department names from your report and adapts, but you should review the flags with domain context. Update the skill with your adjustments.

  • Can I scope to one account or department? Yes — "scan 70xxxx accounts only" or "Marketing department only."

  • Does it post the JE? No — CSV only. Import via NetSuite's JE Import Assistant.

  • Vendor-level analysis? Yes — ad-hoc mode with pivots: ["Department", "Vendor"].

  • Audit trail? Line memos reference the original JE number when the anomaly traces to a specific entry.

  • What period does it scan? The most recent closed period by default. You can point it at another period if needed.

Download & Install

Download the full Numeric Toolkit (15 community skills, including this one):

🔗 numeric-mcp-toolkit.zip

Install via Claude Code CLI:

/plugin marketplace add geoff-lew/numeric-toolkit
/plugin install numeric-mcp-toolkit

Direct link to the Skill

🔗 dept-anomaly-scan.skill

In Cowork, go to Customize in the left sidebar, click +, select the Skills tab, and click Install to upload the downloaded skill file.

Need help? Reach out to support@numeric.io.