# Lead scoring model builder Human Guide

## What This Is For
Builds a custom lead scoring model for a business. It gives the agent a clearer input/output frame for paid media: what context to ask for, what decisions to make, and what usable artifact to return.

Use this as a human-readable version of the Lead scoring model builder agent skill. It is meant for marketers, operators, founders, and other non-coders who want the workflow without reading agent-specific implementation instructions.

## When To Use This
- Use this when you need a repeatable process for lead scoring model builder.
- Use this when the task needs judgment, examples, constraints, or a clear output format rather than a one-off prompt.
- Use this when you want to hand an AI assistant enough context to produce a usable marketing artifact.

## When Not To Use This
- Do not use this when you only need a quick factual answer.
- Do not use this when the work depends on private data you cannot share with the assistant.
- Do not use this as a replacement for legal, compliance, financial, or medical review.

## What You Need Before Starting
- The goal or business outcome you want.
- The audience, customer segment, or market context.
- Any source material the assistant should respect, such as notes, briefs, examples, URLs, or brand guidance.
- Constraints such as tone, length, channel, deadline, region, or approval requirements.
- A clear definition of what a good final answer should look like.

## Step-By-Step Workflow
1. State the job clearly: "Use the Lead scoring model builder guide to help me with..."
2. Add context: audience, goal, offer, channel, source material, and constraints.
3. Ask the assistant to identify missing inputs before producing the final output.
4. Have the assistant follow the skill-specific guidance below.
5. Review the result against the final checklist and ask for revisions where needed.

## Skill-Specific Guidance
- `references/inputs.md` — required, recommended, and optional inputs; the six-step analysis process; batch scoring mode; best practices; trigger phrases and example.
- `references/output-template.md` — the full `lead-scoring-model.md` structure to generate (Sections 1-8, tables, confusion matrix, histogram).
- **Data over intuition.** Trace every point value to a measured lift. If data is insufficient for a dimension, state so explicitly rather than fabricating weights.
- **Simplicity over complexity.** Keep total dimensions to 20-30 signals maximum. A model reps use beats a perfect model they ignore.
- **Continuous calibration.** Build validation and recalibration methodology in from day one; every model degrades over time.
- **No vanity scores.** The model exists to prioritize rep time. If the score does not change rep behavior, it is not useful.
- **Validate.** Hold out 20-30% of historical data, score it, and report precision, recall, F1, AUC-ROC, and a confusion matrix. Analyze false positives and false negatives and iterate.
- **Generate the deliverable.** Write `lead-scoring-model.md` following `references/output-template.md`. Fill every placeholder with data-derived values. Include Section 7 only when a batch of current leads was provided.
- Refuse to build a model on intuition alone. Without historical win/loss data, help the user set up tracking first and revisit in 90 days.
- Show the lift calculation behind every point value.
- Start conservative: under-scoring a few Hot leads beats drowning reps in false positives.
- Never include a signal the CRM cannot reliably capture.

## Decision Points And Nuance
The original skill emphasizes: Contents, Core Principles, Workflow, Guardrails, Supporting file: references/inputs.md, Inputs, Required Inputs, Highly Recommended Inputs, Optional Inputs, Analysis Process.

Use these questions to steer the work:
- What is the intended audience or buyer?
- What source material must be preserved?
- What should the assistant optimize for: clarity, persuasion, accuracy, speed, creativity, or conversion?
- What examples represent the desired quality bar?
- What should the assistant avoid?

## Common Mistakes
- `references/inputs.md` — required, recommended, and optional inputs; the six-step analysis process; batch scoring mode; best practices; trigger phrases and example.
- Never include a signal the CRM cannot reliably capture.
- Follow this sequence rigorously. Do not skip steps.
- Document which attributes do NOT correlate with winning (these are often surprising)
- Hold out 20-30% of historical data for validation (do not use for model building)
- **Demand Data**: Do not build a scoring model on vibes. If the user lacks historical win/loss data, help them set up tracking first and revisit in 90 days.
- **Align With Sales**: The model must make sense to reps. If a rep looks at a "Hot" lead and says "this is obviously not a real opportunity," the model has a credibility problem regardless of what the math says.
- [Limitation 1: e.g., "Model is trained on deals that reached opportunity stage; does not account for leads that never converted to opportunity"]

## Copy-And-Paste Prompt
```text
Use the Lead scoring model builder human guide.

My goal:
[Describe the business outcome]

Audience:
[Describe who this is for]

Context and source material:
[Paste notes, examples, links, or existing copy]

Constraints:
[Tone, length, channel, timeline, must-include items, must-avoid items]

Before producing the final output, ask me for any missing information that would materially improve the result.
```

## Final Checklist
- [ ] The output matches the original goal.
- [ ] The audience and context are reflected in the answer.
- [ ] Important constraints and source material were preserved.
- [ ] The assistant made the relevant decisions explicit.
- [ ] The final artifact is ready to use, review, or hand to the next person.

## Source
This guide was generated from the onewave-ai/claude-skills skill entry for `lead-scoring-model`.

## Source Skill Notes
These notes preserve the nuance from the original skill. Use them as supporting reference when the workflow above feels too generic.

# Lead Scoring Model Builder

Build a data-driven, custom lead scoring model calibrated to actual win/loss history, not generic best practices. Act as a revenue operations analyst and data scientist: every point value must trace to a correlation in the data, and the model must be simple enough that reps actually use it.

## Contents

- `references/inputs.md` — required, recommended, and optional inputs; the six-step analysis process; batch scoring mode; best practices; trigger phrases and example.
- `references/output-template.md` — the full `lead-scoring-model.md` structure to generate (Sections 1-8, tables, confusion matrix, histogram).

## Core Principles

- **Data over intuition.** Trace every point value to a measured lift. If data is insufficient for a dimension, state so explicitly rather than fabricating weights.
- **Simplicity over complexity.** Keep total dimensions to 20-30 signals maximum. A model reps use beats a perfect model they ignore.
- **Continuous calibration.** Build validation and recalibration methodology in from day one; every model degrades over time.
- **No vanity scores.** The model exists to prioritize rep time. If the score does not change rep behavior, it is not useful.

## Workflow

1. **Gather inputs.** Request ICP definition, historical win/loss data (50+ closed deals minimum, 200+ preferred), and a CRM export of current leads. Accept whatever subset is available and note gaps and their accuracy impact. See `references/inputs.md` for the full input checklist.
2. **Run the analysis process.** Execute the six steps in order: data audit, win/loss pattern analysis, dimension construction, threshold calibration, validation, and implementation planning. Do not skip steps. See `references/inputs.md` for the detailed procedure.
3. **Build the four-dimension model.** Construct Firmographic Fit, Behavioral Signals, Engagement Depth, and Intent Indicators, plus negative signals. Assign point values proportional to measured lift and cap each dimension so no single factor dominates.
4. **Calibrate thresholds.** Plot won vs. lost score distributions, find the separation point, and define Hot/Warm/Cool/Cold tiers with expected conversion rates, SLAs, and volumes. Keep Hot small enough to work fully; keep Cold large enough to save rep time.
5. **Validate.** Hold out 20-30% of historical data, score it, and report precision, recall, F1, AUC-ROC, and a confusion matrix. Analyze false positives and false negatives and iterate.
6. **Generate the deliverable.** Write `lead-scoring-model.md` following `references/output-template.md`. Fill every placeholder with data-derived values. Include Section 7 only when a batch of current leads was provided.
7. **Score current leads (when provided).** Load the model, map fields, score each lead, assign tiers, and produce the Section 7 tables ranked by score with recommended actions. See the Batch Scoring Mode in `references/inputs.md`.

## Guardrails

- Refuse to build a model on intuition alone. Without historical win/loss data, help the user set up tracking first and revisit in 90 days.
- Show the lift calculation behind every point value.
- Start conservative: under-scoring a few Hot leads beats drowning reps in false positives.
- Never include a signal the CRM cannot reliably capture.
- Insist on holdout validation before any model goes live.

---

## Supporting file: references/inputs.md

# Inputs and Analysis Process

## Inputs

Request the following inputs. Work with whatever subset is available, but note gaps and their impact on model accuracy.

## Required Inputs

1. **ICP Definition**: Target company profile (industry, size, geography, tech stack, budget range, use case)
2. **Historical Win/Loss Data**: Closed-won and closed-lost deals from the last 12-24 months. Minimum 50 closed deals for statistical relevance; 200+ preferred. Fields needed:
   - Company name, industry, employee count, revenue range
   - Lead source, initial engagement type
   - Deal size, sales cycle length, outcome (won/lost)
   - Loss reason (if lost)
   - Number of touches, stakeholders involved
3. **CRM Export of Current Leads/Opportunities**: The leads to be scored or the pipeline to validate the model against

## Highly Recommended Inputs

4. **Engagement Data**: Email opens, click rates, content downloads, webinar attendance, website visits, demo requests
5. **Firmographic Enrichment**: Tech stack data, funding history, hiring signals, growth rate
6. **Sales Activity Logs**: Call notes, meeting counts, response times, multi-threading depth

## Optional Inputs

7. **Marketing Attribution Data**: First touch, last touch, multi-touch attribution
8. **Intent Data**: Third-party intent signals (Bombora, G2, TrustRadius searches)
9. **Competitive Intelligence**: Which competitors appeared in won vs. lost deals

## Analysis Process

Follow this sequence rigorously. Do not skip steps.

### Step 1: Data Audit

- Inventory all fields available across the provided data
- Identify missing fields and their impact on model completeness
- Check data quality: completeness rates, obvious errors, duplicates
- Flag any survivorship bias (e.g., only seeing leads that made it to opportunity stage)
- Determine sample size adequacy for each dimension
- Document data limitations clearly