Return My TimeBook a Discovery Call

Revenue Operations

Build a Lead-Scoring Rubric

Build an evidence-based lead-scoring rubric whose weights, source fields, uncertainty, and human review points remain inspectable.

By Nick Spisak

Quick facts

Best for
Owners · Founder-operators · Sales leaders
Prompt type
Effectiveness
Expected result
A weighted evidence rubric with source fields, score rationale, explicit unknown and insufficient-data states, validation checks, and human approval gates.
Time saved
60-90 minutes per rubric draft
Required inputs
IDEAL CUSTOMER CRITERIA · DISQUALIFIERS · OBSERVED LEAD FIELDS · SCORING WEIGHTS · QUALIFICATION OWNER
Works with
ChatGPT + Claude

What this prompt does

  • Maps approved fit criteria to weighted, observable evidence instead of intuition.
  • Produces a score explanation while preserving an insufficient-data outcome when evidence is absent.

When to use

  • When a service business wants a transparent scoring draft to compare with human qualification decisions before any automation.

When not to use

  • Do not use it for automatic rejection, live prioritization, CRM updates, or lead contact without validation and named human approval.

The prompt

#CONTEXT:
Draft a weighted lead-scoring rubric for a service business using only supplied criteria, weights, and observed evidence. The result is an analysis for validation, not an automated qualification or adverse decision.

#INPUTS:
- Approved ideal-customer criteria: [IDEAL CUSTOMER CRITERIA]
- Approved disqualifiers: [DISQUALIFIERS]
- Observed lead fields and field provenance: [OBSERVED LEAD FIELDS]
- Proposed criterion weights and scale: [SCORING WEIGHTS]
- Named human qualification owner: [QUALIFICATION OWNER]

#INSTRUCTIONS:
1. Check that every weight maps to one supplied, observable business criterion; report gaps, overlaps, and totals.
2. For each criterion, define the evidence needed for full, partial, zero, and unknown credit.
3. Score only supplied evidence. Do not guess, invent, infer, or fabricate a lead fact.
4. Mark the total "Insufficient data" when missing evidence prevents a defensible score; do not convert unknowns to zero.
5. Do not qualify, disqualify, score, or route a person based on a protected trait or a proxy for a protected trait. Flag suspected proxies for removal and human review.
6. Keep verified hard disqualifiers separate from weighted fit signals and from missing evidence.
7. Recommend validation checks for false positives, false negatives, and uneven outcomes before adoption.
8. Do not update a CRM, prioritize a live queue, route a lead, contact anyone, or send an external message.
9. A named human owner must review and approve every qualification, disqualification, scoring, or routing rule change before operational use.

#RESPONSE FORMAT:
## Weight audit
- Weight total — overlaps — uncovered criteria

## Weighted evidence rubric
| Criterion | Weight | Evidence source | Full / partial / zero evidence | Unknown handling | Score rationale |
|---|---:|---|---|---|---|

## Result states
- Evidence-supported range
- Unknown or Insufficient data — missing evidence
- Verified disqualifier — supplied rule and evidence

## Validation and approval checklist
- Proxy review
- Outcome review
- Named human approval before operational use

Input checklist

  • IDEAL CUSTOMER CRITERIA
  • DISQUALIFIERS
  • OBSERVED LEAD FIELDS
  • SCORING WEIGHTS
  • QUALIFICATION OWNER

Example input

Fictional example — IDEAL CUSTOMER CRITERIA: Commercial property manager, location inside the published service map, and a request for recurring maintenance. DISQUALIFIERS: Verified residential-only request or confirmed location outside the service map. OBSERVED LEAD FIELDS: Company: Willow & Finch Properties; portfolio type: commercial; requested service: recurring maintenance; site address: blank. SCORING WEIGHTS: Commercial portfolio 40, in-area site 35, recurring need 25. QUALIFICATION OWNER: Elena, Revenue Operations Manager.

Expected output structure

  • A weighted evidence rubric with source fields, score rationale, explicit unknown and insufficient-data states, validation checks, and human approval gates.

Customize this prompt

  • Use a small number of observable criteria before considering additional weight complexity.
  • Compare rubric recommendations with reviewed historical decisions before any operational use.

Guardrails

  • Do not qualify, disqualify, score, route, or recommend an adverse decision based on a protected trait or any proxy for a protected trait.
  • Do not invent missing lead facts or silently score unknown evidence as zero; use Insufficient data.
  • A named human owner must review and approve every qualification, disqualification, scoring, routing rule change, policy exception, account update, external message, price, promise, permission, and legal commitment before any operational use.
  • The assistant must not access or change a live CRM, prioritize a live queue, route a lead, contact anyone, or send an external message.

KEEP BUILDING

Related prompts

Related articles

NEXT STEP

Next step

Map where revenue is leaking.