AI Stack for Marketo Lead Scoring

Most Marketo lead scoring models were built once, based on what seemed logical at the time, and never validated against actual closed-won data. The result is a scoring model that rewards the wrong behaviors — high scores for people who download whitepapers and never buy, low scores for buyers who go straight from demo request to close. This stack rebuilds your scoring model from the data that matters: what Marketo behaviors and demographic attributes actually appear in your closed-won cohort. Claude runs the diagnostic, identifies which scoring rules are predictive and which are noise, and outputs the specific changes to make your model surface real buying intent.

The Stack

The Prompt

This stack is built around the Marketo Nurture Diagnostic Prompt. Here's the abbreviated version — the full prompt with all variables and usage notes is on its own page.

You are a marketing operations analyst auditing a Marketo lead scoring model.

Using the current Marketo scoring rules and Salesforce closed-won attribute data below,
diagnose the scoring model and produce:

1. Scoring rules that correlate with closed-won behavior (keep and potentially increase)
2. Scoring rules that don't appear in the closed-won cohort (reduce or remove)
3. Missing behaviors that appear frequently in closed-won but aren't currently scored
4. Recommended demographic score adjustments based on closed-won company attributes
5. Recommended MQL threshold adjustment with rationale
6. Priority order for implementing changes (highest-impact first)

The Workflow

1. Export your current Marketo scoring model

   Pull the complete scoring rule set: which activities score, how many points each earns, the demographic scoring model, and the current MQL threshold. Also export the last 90 days of score activity — which behaviors are driving the most score changes across your database.

2. Export Salesforce closed-won attributes

   Pull closed-won opportunities from the last 12 months: job title, company size, industry, lead source, first touch, and time-in-stage at each funnel step. Export closed-lost for comparison. The difference between closed-won and closed-lost attribute patterns is your scoring signal.

3. Compare current scoring rules against closed-won behavior

   Before running the prompt, manually compare: which Marketo-scored behaviors appear frequently in your closed-won cohort, and which don't? This context tells Claude where to focus the audit rather than starting from scratch.

4. Paste Marketo scoring data and closed-won analysis into the prompt

   Give Claude the current scoring model and the closed-won attribute comparison. Claude identifies predictive rules to keep, noise rules to reduce, behaviors that should be added, and the specific threshold adjustment that will improve MQL quality.

5. Implement the updated model in Marketo and document in Notion

   Start with behavior threshold changes — they're the highest-impact adjustment. Then revise demographic scoring. Document every change in Notion with the rationale so the next ops person understands the logic without reverse-engineering the model.

What This Replaces

• Marketo scoring models built on vendor recommendations instead of your own closed-won data
• MQL thresholds that were set arbitrarily and never calibrated against actual funnel conversion rates
• Scoring audits that require a full-time ops analyst weeks to run because there's no structured diagnostic process

Related Stacks

• AI Stack for Marketo Nurture Audit →
• AI Stack for Inbound Lead Routing →
• AI Stack for Campaign Operations →