Loss Run Analytics

Turn 4–5 hours of manual loss run analysis into 15 minutes — with actuarial-grade domain knowledge.

A free Claude Cowork plugin that serves as an alternative to Applied Epic's loss run analysis workflow for insurance agencies. Imports mixed-format loss run documents (PDF, Excel, CSV), normalizes and cleans the data using ACORD Data Dictionary v3.0 standards, performs trend analysis with actuarial loss development factors, calculates EMod using the NCCI Experience Rating Plan Manual formula, generates professional reports with industry benchmark comparisons, and drafts risk assessments using the IRMI severity-frequency framework.

Important: All documents generated by this plugin are drafts intended as starting points. Insurance documents, analyses, and recommendations must be reviewed by a licensed insurance professional before use with clients or carriers. This plugin does not provide insurance advice.

Installation

1. Download the pf-loss-run-analytics.plugin file
2. Open Claude Desktop and navigate to Settings > Plugins
3. Click Install Plugin and select the downloaded .plugin file
4. The plugin will be installed and available immediately

Note: All data stays local on your machine. No external API calls or cloud storage required.

Why This Exists

Insurance agents spend 4–5 hours per account manually parsing loss run PDFs from carriers, reformatting data, computing frequency/severity trends, estimating EMod factors, and writing up analysis. Applied Epic charges $3,000–$12,000/year for this workflow. This plugin does it in 15 minutes at Included with your subscription — and it produces the same underwriter-ready outputs your team expects, backed by authoritative industry references (NCCI, ACORD, NAIC, CAS, AM Best).

Quick Start

1. Install the plugin in Claude Cowork
2. Run /loss:loss-setup — initializes your workspace folders and configures NCCI class codes
3. Place your loss run files (PDF, Excel, or CSV) in the inbound/ folder
4. Run /loss:loss-pipeline — the full pipeline runs automatically
5. Find your reports in outbound/ — ready to send to underwriters or clients

Commands

Command	Description
/loss:loss-setup	Initialize workspace, check dependencies, configure agency and NCCI class codes
/loss:loss-import [file]	Parse loss run documents (PDF/Excel/CSV) using ACORD Data Dictionary field mapping
/loss:loss-clean	Normalize claims to ACORD/NAIC categories, validate with actuarial reasonability checks
/loss:loss-analyze	Trend analysis with loss development factors (LDFs) and NAIC benchmark comparisons
/loss:loss-emod [state]	Calculate EMod using NCCI Experience Rating Plan Manual formula (Split Point, D-Ratio, Weighting)
/loss:loss-report [insured]	Generate professional DOCX report with actuarial context and methodology citations
/loss:loss-risk [type]	Risk assessment using IRMI framework, declination letter with NAIC compliance guidance
/loss:loss-dashboard	Interactive visual dashboard — "show me"
/loss:loss-pipeline [file]	Run complete pipeline in one command
/loss:loss-status	Show pipeline progress and next step
/loss:loss-resume [stage]	Resume pipeline from last checkpoint

How It Works

The plugin runs a 6-stage pipeline:

inbound/ (your loss run files) → [loss-import] → processing/loss-parsed.json → [loss-clean] → processing/loss-cleaned.json → [loss-analyze] → processing/loss-analysis.json → [loss-emod] → processing/loss-emod.json → [loss-report] → outbound/loss-loss-report.docx + [loss-risk] → outbound/loss-risk-summary.docx

Each stage checkpoints its output. If interrupted, /loss:loss-resume picks up where you left off.

AI-Powered Features

This plugin leverages Claude's AI capabilities combined with authoritative insurance domain knowledge to go beyond simple automation:

• ACORD-Standard Format Parsing — Maps carrier column headers to ACORD Data Dictionary v3.0 standard field names. Recognizes 12 data fields across 80+ column header variations from the top 20 P&C carriers. Detects NAIC cause-of-loss codes and NCCI class codes automatically.
• NAIC-Aligned Claims Classification — Maps free-text claim descriptions to 7 ACORD standard coverage categories (Workers Comp, General Liability, Commercial Auto, Property, Professional Liability/E&O, Umbrella/Excess, Other) with NAIC Annual Statement Schedule P line-of-business codes and ISO form references (CG 00 01, CA 00 01, CP 00 10).
• Actuarial Loss Development — Applies industry-standard loss development factors (LDFs) per CAS Actuarial Standards of Practice No. 43 to estimate ultimate incurred losses and IBNR (Incurred But Not Reported) exposure for immature policy years.
• NAIC Benchmark Comparison — Compares the insured's loss ratios against industry benchmarks from NAIC Annual Statement data / AM Best Aggregates & Averages across 7 industry segments.
• NCCI EMod Calculation — Implements the full NCCI Experience Rating Plan Manual (2024 ed.) Formula F-1 with proper Split Point ($18,500), D-Ratio tables by Hazard Group, Weighting Values, and Ballast calculations. Supports state bureau variations (WCIRB, NYCIRB, MWCIA).
• OSHA-Aligned Recommendations — Generates 6 actionable EMod improvement recommendations targeting specific NCCI formula components, with OSHA publication references and experience period management guidance.
• IRMI Risk Classification — Evaluates risk using the IRMI severity-frequency matrix combined with AM Best rating methodology — a 6-factor weighted scoring system covering EMod, trends, benchmarks, IBNR, and concentration risk.
• Regulatory-Compliant Declinations — Declination letters include NAIC Market Conduct Model Act guidance and state-specific adverse action requirements (NY, CA, TX) to help agencies maintain compliance.
• Executive Report Synthesis — Synthesizes all analysis into a 9-section professional report with actuarial terminology, methodology citations, and proper legal disclaimers.

What It's an Alternative To

SaaS Tool	Annual Cost	Key Overlap
Applied Epic	$3,000–$12,000	Loss run import, trend analysis, EMod
Hawksoft	$2,400–$6,000	Loss run management, reporting
AMS360	$4,000–$10,000	Loss analytics, risk scoring
Excel + manual work	included with your subscription but 4-5 hrs/account	All of the above, manually

Key Differentiators

• Included with your subscription — No licensing, no per-seat fees, no per-report charges
• Actuarial-grade analysis — Uses the same NCCI formulas, ACORD standards, and CAS methodology that professional actuaries apply
• No carrier API required — Works with exported files from any carrier portal
• Privacy by default — data is processed within the Cowork environment. Processing happens within the Cowork environment.
• Mixed-format support — Handles PDF, Excel, and CSV in the same run (most tools require a single format)
• Checkpoint/resume — If a session is interrupted, resume exactly where you left off
• Source-cited outputs — Every calculation and classification references its authoritative source

Feature Comparison

Feature	This Plugin	Applied Epic	Hawksoft
PDF import	✅ full	✅ full	⚠️ partial
Excel/CSV import	✅ full	✅ full	✅ full
ACORD classification (7 categories)	✅ full	✅ full	⚠️ partial
NCCI EMod calculation (Formula F-1)	✅ full (estimate)	✅ full	❌ none
Loss development factors (LDFs)	✅ industry standard	✅ full	❌ none
IBNR estimation	✅ full	✅ full	❌ none
NAIC benchmark comparison	✅ full	⚠️ partial	❌ none
IRMI risk classification	✅ full	⚠️ partial	❌ none
Trend analysis	✅ full	✅ full	⚠️ partial
DOCX report output (9 sections)	✅ full	✅ full	⚠️ partial
Declination letter (NAIC-compliant)	✅ full	⚠️ partial	❌ none
AI narrative insights	✅ full	❌ none	❌ none
Carrier API integration	❌ none	✅ full	✅ full
Cost	included with your subscription	$3K–$12K/yr	$2.4K–$6K/yr

Estimated Cost per Use

Disclaimer: Token estimates are approximate and based on typical usage patterns measured from skill prompt sizes. Actual costs vary with input data size, conversation length, and complexity. Estimates use Claude Sonnet 4.6 pricing ($3/1M input, $15/1M output). Cowork and Claude Desktop subscription users (Pro/Max/Team) are not charged per-token — these estimates apply only to direct Anthropic API usage. Running stages individually in fresh sessions uses fewer input tokens than running the full pipeline sequentially, because pipeline mode accumulates conversation history across stages.

Per skill (run individually in a fresh session):

Stage	Skill Prompt	User Input	Total Input	Output	Est. Cost
loss-dashboard	~4.4K	~500	~8.1K	~8.0K	~$0.14
loss-clean	~3.8K	~800	~7.8K	~3.8K	~$0.08
loss-emod	~5.2K	~800	~9.4K	~5.2K	~$0.11
loss-analyze	~5.6K	~800	~9.7K	~5.6K	~$0.11
loss-risk	~4.2K	~800	~8.4K	~4.2K	~$0.09
loss-import	~4.3K	~2.0K	~9.6K	~2.0K	~$0.06
loss-report	~4.0K	~800	~8.3K	~6.0K	~$0.11
Standalone total			~61.3K	~34.8K	~$0.71

Full pipeline (all stages in one session — context accumulates):

Stage	Base Input	+ History	Total Input	Output	Est. Cost
loss-dashboard	~9.1K	0	~9.1K	~8.0K	~$0.15
loss-clean	~8.7K	~8.5K	~17.2K	~3.8K	~$0.11
loss-emod	~10.2K	~13.1K	~23.3K	~5.2K	~$0.15
loss-analyze	~10.5K	~19.1K	~29.6K	~5.6K	~$0.17
loss-risk	~9.1K	~25.5K	~34.6K	~4.2K	~$0.17
loss-import	~10.4K	~30.5K	~40.9K	~2.0K	~$0.15
loss-report	~8.9K	~34.5K	~43.5K	~6.0K	~$0.22
Pipeline total			~198.2K	~34.8K	~$1.12

Running the full pipeline once typically costs $0.78–$1.45 in API tokens (Claude Sonnet 4.6).

Known Limitations & Workarounds

This is a Claude Cowork plugin, not a traditional SaaS application. It has architectural constraints you should know about before adopting it.

Limitation	Impact	Workaround
Scanned PDFs not supported	Cannot extract text from image-based PDFs	Request text-based PDFs from carriers, or export loss run to Excel before importing
EMod is an estimate, not actuarial opinion	Uses NCCI standard formula with industry LDFs; not a certified actuarial calculation per CAS ASOP	Verify against your NCCI bureau or carrier's official EMod worksheet. Output includes full formula documentation for audit.
State bureau variations	CA (WCIRB), NY (NYCIRB) use different formulas	Plugin flags non-NCCI states and recommends verification with the appropriate state bureau
LDFs are industry averages	Loss development factors from NAIC aggregate data, not insured-specific	For more accurate development, obtain insured's own loss triangle from carrier. Plugin notes estimation uncertainty per ASOP No. 43.
No carrier API connectivity	Cannot pull loss runs automatically from carrier portals	Export loss runs from carrier portals as CSV/Excel; place in inbound/
No persistent database	Each session processes the files in your inbound/ folder; no multi-account database	Organize folders by client/date. All outputs saved as versioned files in outbound/ and archive/.
Session-based processing	Pipeline runs within a single Claude session	Processing data persists in local files between sessions. Use /loss:loss-resume to continue.

Context & Performance Guide

Session Management

Claude's context window (200K tokens) is shared across the full session. For the best output quality, start a fresh Claude session before running the full pipeline. If you've been chatting extensively before running the pipeline, a new session gives the plugin more room to produce high-quality narrative insights and complete reports.

Data Volume Guidance

Tested with up to 500 claims in a single run. For larger datasets, the plugin automatically switches to Python aggregation for statistical computations — Claude only interprets the pre-computed aggregates, not individual claims. This keeps analysis quality high regardless of dataset size.

For multi-account batches: run each account in a separate session for cleanest results.

Tips for Best Results

1. Provide input data as files rather than pasting into chat — file reading is more context-efficient
2. Keep your messages concise during pipeline runs — every message stays in the context window
3. For the highest quality output on large datasets, run each stage individually using stage-specific commands
4. If output quality degrades (shorter narratives, missing sections), start a new session and run /loss:loss-resume to continue

Degradation Signals

If responses become shorter, less detailed, or skip sections that previously worked, start a new session and re-run the current stage. The plugin saves all intermediate state — you won't lose progress.

Authoritative Sources

This plugin references the following industry standards and regulatory frameworks:

• NCCI Experience Rating Plan Manual (2024 ed.) — EMod formula, Split Point, D-Ratio tables, Weighting Values
• NCCI Basic Manual for Workers Compensation — Class codes, Expected Loss Rates, Hazard Groups
• ACORD Data Dictionary v3.0 — Standardized field names and coverage codes
• NAIC Property Casualty Insurance Statistical Plan — Claim coding, cause of loss codes
• CAS Actuarial Standards of Practice No. 43 — Unpaid claim estimation methodology (LDFs)
• AM Best Aggregates & Averages — Industry benchmark loss ratios
• IRMI (International Risk Management Institute) — Risk classification framework
• OSHA Recommended Practices — Workplace safety and reporting standards
• ISO Commercial Forms — Coverage form references (CG 00 01, CA 00 01, CP 00 10)
• NAIC Market Conduct Model Act — Declination letter compliance

Requirements

• Claude Desktop with Cowork mode enabled
• Python 3.8+ (for document generation scripts)
• pip packages: pdfplumber, openpyxl, python-docx (auto-installed by /loss:loss-setup)

Important Disclaimers

• AI-Generated Content: This plugin uses AI (LLM) technology which can produce inaccurate or incomplete outputs. All content should be treated as a starting point and reviewed for accuracy before use.
• Not Professional Advice: Outputs do not constitute legal, financial, tax, medical, or other professional advice. Consult qualified professionals before making decisions based on generated content.
• No Compliance Guarantee: References to industry standards, regulations, or guidelines are for informational purposes only. This plugin does not guarantee compliance with any law or regulation. Users are responsible for verifying all outputs meet their specific regulatory requirements.
• No Endorsement or Affiliation: Mention of third-party products, standards, or organizations does not imply endorsement, partnership, or certification by those entities.
• Not Financial or Tax Advice: Financial calculations, tax estimates, and related outputs are for informational purposes only. Consult a qualified CPA or financial advisor for tax and financial decisions.