Impact Sizing Skill

Purpose: Guide users through a structured impact sizing methodology to translate product hypotheses into ARR estimates, then generate a complete Excel workbook with all calculations.

When to Use This Skill

• When a PRD is in the PROPOSED stage
• Before significant Product & Design effort has been invested
• When comparing product opportunities for prioritisation
• When you need to justify why something should be built (or not)

Data Collection Approach

This skill collects data through a flexible, multi-format approach:

Accepted Input Methods

1. Direct answers - User types numbers or descriptions in response to questions
2. File uploads - CSV exports, Excel files, or screenshots from Salesforce/Twine/Productboard
3. Pasted text - Copy/paste from Confluence tables, Slack threads, or reports
4. Partial data - Work with whatever data is available; estimate or skip missing sections

How Claude Handles Different Formats

For Salesforce exports (CSV/Excel):

• Claude will read the file and extract relevant columns (Account Name, ARR, Closed Lost Reason, etc.)
• Filter for feature-related keywords automatically
• Sum values and identify relevant accounts

For Confluence/Notion tables:

• User can paste the table text directly
• Claude will parse the structure and extract data

For Twine/Slack:

• User provides account names and ARR values mentioned
• Claude will help structure the data

For screenshots:

• Claude can read text from screenshots if data isn't exportable
• User should confirm extracted values

When Data Isn't Available

If the user doesn't have access to certain data sources, Claude should:

1. Note the gap in the assumptions sheet
2. Provide a conservative estimate based on available data
3. Mark confidence as "Low" for that category
4. Continue with the sizing using what's available

Phase 1: Discovery Questions

When this skill is invoked, Claude MUST ask the following questions to gather inputs. Use the AskUserQuestion tool with clear, structured questions.

1.1 Core Hypothesis

Ask the user to provide their hypothesis in the standard format:

"By <making a change> we will see <metric> <increase/decrease/stay same> because <customer problem solved>"

The metric should be one of the Big 5 metrics:

• Bookings (GTM)
• Pipeline Created (GTM)
• Gross Dollar Retention / GDR (GTM)
• Users (Product)
• MAU (Product)

1.2 Feature Details

• Feature name: What is the feature or capability being sized?
• Product area: Which product area does this belong to? (e.g., Capture, Reporting, Assets, Training, etc.)
• Is this new functionality or enhancing existing?: Determines whether to use top-down or bottom-up TAM/SAM/SOM

1.3 Strategic Focus

Confirm alignment with your company's strategic focus:

• Priority Industries: [Your target verticals, e.g. Manufacturing, Retail, Healthcare]
• Priority Geographies: [Your target regions, e.g. USA, UK, ANZ]
• Customer Segment: [Your target segments, e.g. Enterprise (>1,000 employees), Mid-market]

1.4 Data Sources

Ask what data the user has access to. For each source, explain how to get the data:

CRM - Closed Lost/Won:

Export from your CRM's closed sales report (e.g. Salesforce, HubSpot)

• Export as CSV with deal stage, close reason, and amount columns
• Or just tell me the feature-related closed lost value if you've already looked it up

Feedback Tool - Feature Requests:

Go to your feature in your feedback tool (e.g. Productboard, Canny, Intercom) and note:

• Number of requests for this specific feature
• Total requests in the product area (for calculating proportion)

Customer Intelligence - Churn Risk:

Search your customer intelligence tool (e.g. Twine, Gong, Chorus) for keywords related to your feature

• List customer names and their ARR from accounts expressing frustration
• Or upload the report export

Churned Customers:

Check your churn analysis data or spreadsheet

• Find customers who churned citing this capability
• List their names and ARR values

Current Pipeline:

Any deals currently in pipeline that depend on this feature?

• Usually surfaced directly by GTM team

Phase 1.5: Processing Uploaded Data

When the user uploads files or pastes data, Claude should process it as follows:

Salesforce CSV/Excel Exports

Expected columns to look for:

• Account Name or Opportunity Name
• Stage (filter for "Closed Lost" or "Closed Won")
• Closed Lost Reason or Closed Lost/Churn Reason
• Additional Details (search for feature keywords here)
• Reporting Total Price (converted) or Amount or ARR

Processing steps:

1. Read the file using pandas
2. Filter Stage = "Closed Lost" and Reason contains "Product"
3. Search Additional Details for feature-related keywords
4. Sum the ARR/Amount column for matching rows
5. Present findings to user for confirmation

Example code Claude should use:

import pandas as pd

df = pd.read_csv('salesforce_export.csv')  # or read_excel

# Filter for product-related closed lost
closed_lost = df[
    (df['Stage'].str.contains('Closed Lost', case=False, na=False)) &
    (df['Closed Lost Reason'].str.contains('Product', case=False, na=False))
]

# Search for feature keywords
keywords = ['asset', 'multi-asset', 'multiple assets']  # customize per feature
feature_related = closed_lost[
    closed_lost['Additional Details'].str.contains('|'.join(keywords), case=False, na=False)
]

total_value = feature_related['Reporting Total Price (converted)'].sum()
print(f"Found {len(feature_related)} closed-lost deals totaling ${total_value:,.0f}")

Twine Exports

Expected format: Usually JSON or CSV with call summaries Look for: Customer name, ARR (may need to cross-reference with Salesforce), key quotes about frustration

Productboard Data

Expected format: Feature request counts, often from the Insights/Trends view Look for: Total request count for the feature, total requests in product area

Pasted Confluence/Slack Text

Processing approach:

1. Look for table structures (pipe-delimited or tab-delimited)
2. Extract customer names and numeric values
3. Ask user to confirm which column is ARR

Phase 2: TAM/SAM/SOM Estimation

Based on whether this is NEW functionality or ENHANCEMENT to existing:

For NEW Functionality (Top-Down Approach)

Use AI (Perplexity/Claude) to estimate market size with prompts like:

TAM Prompt Template:

Estimate the Total Addressable Market (TAM) for a software tool that [DESCRIBE CAPABILITY]:
- TAM methodology: Top-down from global spend or relevant software categories
- List all key assumptions and provide sensitivity ranges (low / base / high)
- Output: Provide TAM in AUD (annual), with low/base/high scenarios
- Sources & citations: Use recent sources, cite at least 5 reputable sources

SAM Prompt Template (follow-up):

Turn this into a SAM estimate with the following focus areas:
- Industry focus: manufacturing, retail, QSR, and hospitality (but don't exclude others)
- Geographical focus: United States, Canada, APAC, and English-speaking EMEA
- Customer focus: Enterprise customers, but don't preclude SMBs
- List all key assumptions and provide sensitivity ranges (low / base / high)
- Output: Provide SAM in AUD (annual), with low/base/high scenarios

SOM Calculation:

• Estimate target customers in regions/industries
• Apply realistic capture rate (typically 5%)
• Apply average seat count (typically 5 seats)
• Apply seat price ($24/month or $288/year standard, or $367/year for enterprise blend)
• Apply win rate (23%) and deal cycle adjustment

For ENHANCEMENT (Bottom-Up Approach)

Use internal data sources:

TAM: Use Industry Playbooks + All Customers Report to estimate total addressable orgs SAM: Use Productboard feature requests as % of total requests in that product area SOM: Apply 5% capture rate × average seats × price × win rate

Phase 3: Churn Analysis

3.1 Churned Customers

Search for customers who churned specifically due to this capability missing:

• Data Source: Your product churn analysis spreadsheet
• Data Source: Your CRM churn/renewal reports
• Look for keywords related to the feature in churn reasons
• Sum the ARR of churned accounts

3.2 Tangible Churn Risk

Identify customers at risk of churning if capability isn't delivered:

• Data Source: Twine - search for feature-related frustration
• Data Source: #account-risks Slack channel
• Cross-reference with Salesforce account ARR values

3.3 Retention Uplift

Calculate expected year-on-year churn reduction:

Retention Uplift = (Churned Customers ARR / Total Prior Year Churn) × GDR Uplift Target

• Prior Year Total Company Churn: [Your churn figure]
• GDR Uplift Target: [Your target, e.g. 2% reduction]

Phase 4: Opportunities Analysis

4.1 Closed Won Opportunity (from Closed Lost data)

• Data Source: Your CRM closed sales report (last 2 fiscal years)
• Filter: Stage = "Closed Lost", Reason = Product-related
• Search Additional Details for feature-related keywords
• Calculate: Feature-related Closed Lost / Total Closed Deals × Next Year Bookings Target

4.2 Current Opportunities

• Any pipeline opportunities specifically dependent on this feature
• Usually surfaced directly by GTM team

4.3 Other Conversion Opportunities

• Free-to-paid conversion impact
• Relevant for features in onboarding or early user journey

Phase 5: Generate Excel Workbook

After gathering all inputs, generate an Excel workbook with the following structure:

Sheet 1: Summary

Sheet 2: Detailed Calculations

Include all inputs, assumptions, formulas with cell references, and source citations.

Sheet 3: Data Sources

Table of all data sources used with links and access dates.

Key Reference Data (Customise for Your Company)

Key Resource Links (Customise for Your Company)

Excel Generation Instructions

When generating the Excel file, Claude should:

1. Use the xlsx skill for proper Excel creation
2. Include formulas (not hardcoded calculations) so the workbook is dynamic
3. Apply proper formatting:
   • Blue text for user inputs
   • Black text for formulas
   • Yellow highlight for key assumptions
   • Currency format for ARR figures
4. Include source citations in adjacent cells
5. Add a hypothesis header at the top of the main sheet
6. Calculate quarterly and annualized figures where applicable
7. Run recalc.py to ensure all formulas compute correctly

Example Output

For a feature like "Multiple Assets in Inspections", the final impact sizing came to $800K ARR broken down as:

• Top-down SOM: $750K
• Bottom-up SOM: $135K
• Churned Customers: $25K
• Tangible Churn Risk: $570K
• Retention Uplift: $24K
• Closed Won Opportunity: $49K

(Note: Not all categories are additive - some overlap. Use judgment on which to include in total.)

Workflow Summary

1. User invokes /impact-sizing skill
2. Claude asks discovery questions (hypothesis, feature details, strategic alignment)
3. Claude guides TAM/SAM/SOM estimation (top-down or bottom-up based on feature type)
4. Claude asks for churn data inputs
5. Claude asks for opportunities data inputs
6. Claude generates comprehensive Excel workbook
7. Claude provides summary and key findings

Data Sensitivity

This skill produces outputs containing customer names, deal values, ARR figures, churn reasons, and pipeline data — all commercially sensitive.

• Anonymise where possible. Use company identifiers (e.g., "Account A") in the workbook unless the user needs real names for internal use.
• Store securely. The Excel workbook should not be saved to shared/public directories or committed to git repositories.
• CRM data is confidential. Salesforce exports, churn analysis, and pipeline data pulled during this workflow are production data — treat accordingly.
• Warn before sharing. If the user asks to post results to Slack or Confluence, confirm they're comfortable with the audience seeing customer-level financial data.

Important Notes

• Cite all data sources - transparency is critical
• Make assumptions explicit - it's okay to estimate, just be clear about it
• Apply confidence levels when appropriate (High/Medium/Low)
• Sanity check results - if a number seems unrealistic, revisit assumptions
• The goal isn't perfection - it's consistent comparison across opportunities