Marketing Attribution Agent: Cross-Channel Insights from CRM + Product Data

Build an agent that tracks marketing performance across channels by combining CRM data with product usage, giving you true attribution insights without the manual work.

Why Build a Marketing Attribution Agent?

Most marketing teams struggle to understand which channels actually drive conversions. You have data in Google Ads, Facebook Ads, email platforms, your CRM, and product analytics—but connecting the dots requires hours of manual work. In this guide, we'll build a marketing attribution agent that automatically tracks customer journeys, attributes conversions to channels, and surfaces insights across your entire marketing stack.

What This Agent Is Expected To Do

This agent answers concrete questions about marketing performance and attribution. Here's what it should handle:

• "Which channels drove the most conversions this quarter?"
• "What's the customer journey for customers who converted from our Q1 campaign?"
• "How do first-touch and last-touch attribution compare for our email campaigns?"
• "Which marketing channels have the best ROI when you include product usage data?"
• "What's the attribution breakdown for customers who became power users?"
• "Which campaigns are driving customers who actually use the product?"

The agent doesn't just return channel metrics—it provides attribution analysis, customer journey insights, and ROI calculations across your entire marketing stack.

Who This Agent Is For

This agent is built for teams that need marketing attribution across multiple systems:

• Marketing Teams – Channel performance, campaign ROI, attribution analysis without manual spreadsheets
• Marketing Ops – Attribution modeling, channel optimization, budget allocation insights
• Growth Teams – Customer journey analysis, conversion path optimization, channel mix insights
• Leadership – Marketing ROI visibility, channel effectiveness, budget allocation decisions

If you live in marketing platforms, CRM, product analytics, and attribution tools, this agent is for you.

Core Use Cases

Here are the scenarios where this agent shines:

True attribution without manual work: Instead of spending hours exporting data from Google Ads, Facebook, email, CRM, and product analytics, then manually matching conversions, the agent gives you complete attribution analysis—first touch, last touch, linear, and custom models—in one question.

Customer journey visibility: The agent shows the complete customer journey from first touchpoint to conversion, including which channels they interacted with, when, and how that influenced their decision. You can see what actually works.

ROI calculations that include product usage: Most attribution tools stop at conversion. This agent includes product usage data, so you can see which channels drive customers who actually use your product, not just sign up.

Channel optimization insights: The agent surfaces which channels work best for which customer segments, which campaigns drive high-value customers, and where to allocate budget for maximum impact.

The Ingredients: Data You'll Need

To build this agent, you'll need data from several systems:

Marketing channel data (from Google Ads, Facebook Ads, LinkedIn, etc.):

• Campaigns: campaign name, channel, spend, impressions, clicks, conversions
• Ad groups: ad group name, targeting, performance metrics
• Keywords: keyword performance, search terms, cost per click

Email marketing data (from email platforms):

• Campaigns: campaign name, send date, open rate, click rate, conversion rate
• Recipients: email, segment, engagement history
• Performance: revenue attributed, ROI

CRM data (from HubSpot, Salesforce, etc.):

• Contacts: contact source, first touchpoint, lead source, conversion date
• Deals: deal source, campaign attribution, revenue
• Activities: email opens, website visits, form submissions

Product usage data (from product analytics):

• Events: signup events, feature usage, activation events
• Users: user ID, signup date, activation date, feature adoption
• Engagement: daily active users, feature usage frequency, power user status

Website analytics data (from Google Analytics, Segment, etc.):

• Sessions: session source, medium, campaign, landing page
• Events: conversion events, goal completions, user actions
• Attribution: first touch, last touch, assisted conversions

Derived fields (calculated in the view):

• customer_journey: Complete sequence of touchpoints from first to conversion
• first_touch_channel: Channel of first touchpoint
• last_touch_channel: Channel of last touchpoint before conversion
• attribution_score: Calculated attribution score based on model (first touch, last touch, linear, time decay)
• roi_with_usage: ROI calculation that includes product usage value
• power_user_rate: Percentage of customers from channel who became power users

Here's an example SQL view:

-- Marketing Attribution View
SELECT 
  -- Customer/Contact data from CRM
  c.contact_id,
  c.email,
  c.company_name,
  c.first_touch_date,
  c.first_touch_channel,
  c.conversion_date,
  c.conversion_source,
  
  -- Marketing touchpoints
  COUNT(DISTINCT t.touchpoint_id) as total_touchpoints,
  MIN(t.touchpoint_date) as first_touchpoint_date,
  MAX(t.touchpoint_date) as last_touchpoint_date,
  STRING_AGG(DISTINCT t.channel ORDER BY t.touchpoint_date, ', ') as journey_channels,
  
  -- Channel performance
  COUNT(DISTINCT CASE WHEN t.channel = 'google_ads' THEN t.touchpoint_id END) as google_ads_touchpoints,
  COUNT(DISTINCT CASE WHEN t.channel = 'facebook_ads' THEN t.touchpoint_id END) as facebook_ads_touchpoints,
  COUNT(DISTINCT CASE WHEN t.channel = 'email' THEN t.touchpoint_id END) as email_touchpoints,
  COUNT(DISTINCT CASE WHEN t.channel = 'linkedin_ads' THEN t.touchpoint_id END) as linkedin_ads_touchpoints,
  COUNT(DISTINCT CASE WHEN t.channel = 'organic' THEN t.touchpoint_id END) as organic_touchpoints,
  
  -- Campaign data
  SUM(DISTINCT ca.spend) as total_marketing_spend,
  SUM(DISTINCT ca.clicks) as total_clicks,
  SUM(DISTINCT ca.conversions) as total_conversions,
  
  -- Product usage context
  u.signup_date,
  u.activation_date,
  DATEDIFF(u.activation_date, u.signup_date) as days_to_activate,
  u.daily_active_users_30d,
  u.feature_adoption_score,
  CASE 
    WHEN u.daily_active_users_30d > 20 AND u.feature_adoption_score > 0.7 THEN 'power_user'
    WHEN u.daily_active_users_30d > 10 AND u.feature_adoption_score > 0.5 THEN 'active_user'
    ELSE 'inactive_user'
  END as user_segment,
  
  -- Revenue data
  d.deal_amount,
  d.deal_status,
  s.mrr,
  s.subscription_status,
  
  -- Attribution calculations
  MIN(CASE WHEN t.touchpoint_date = MIN(t.touchpoint_date) OVER (PARTITION BY c.contact_id) THEN t.channel END) as first_touch_channel_calc,
  MAX(CASE WHEN t.touchpoint_date = MAX(t.touchpoint_date) OVER (PARTITION BY c.contact_id) THEN t.channel END) as last_touch_channel_calc,
  CASE 
    WHEN COUNT(DISTINCT t.touchpoint_id) = 1 THEN MIN(t.channel)
    ELSE 'multi_touch'
  END as attribution_type,
  
  -- ROI calculations
  CASE 
    WHEN SUM(DISTINCT ca.spend) > 0 THEN (COALESCE(d.deal_amount, s.mrr * 12) - SUM(DISTINCT ca.spend)) / SUM(DISTINCT ca.spend) * 100
    ELSE NULL
  END as roi_percent,
  CASE 
    WHEN SUM(DISTINCT ca.spend) > 0 AND u.user_segment = 'power_user' THEN (COALESCE(d.deal_amount, s.mrr * 12) * 1.5 - SUM(DISTINCT ca.spend)) / SUM(DISTINCT ca.spend) * 100
    ELSE NULL
  END as roi_with_usage_value
  
FROM hubspot.contacts c
LEFT JOIN marketing.touchpoints t ON c.contact_id = t.contact_id
LEFT JOIN marketing.campaigns ca ON t.campaign_id = ca.campaign_id
LEFT JOIN product_analytics.users u ON c.email = u.email
LEFT JOIN hubspot.deals d ON c.contact_id = d.contact_id
LEFT JOIN stripe.subscriptions s ON c.email = s.customer_email
WHERE c.conversion_date >= DATE_SUB(CURRENT_DATE(), INTERVAL 90 DAY)
GROUP BY 
  c.contact_id, c.email, c.company_name, c.first_touch_date, c.first_touch_channel,
  c.conversion_date, c.conversion_source,
  u.signup_date, u.activation_date, u.daily_active_users_30d, u.feature_adoption_score,
  d.deal_amount, d.deal_status, s.mrr, s.subscription_status;

This view joins data from your CRM (contacts, deals), marketing platforms (campaigns, touchpoints), product analytics (usage, activation), and billing (revenue), and calculates attribution metrics and ROI.

Key points:

• The view joins across multiple systems in a single query
• It calculates derived fields (first touch, last touch, attribution type, ROI)
• It filters to recent conversions (last 90 days)
• It includes product usage data to calculate true ROI

Step 4: Add MCP Tools on Top of This View

Create MCP tools that expose this view to agents. Tools translate natural language questions into SQL queries.

Tool 1: get_channel_attribution(model: string, period: string)

This tool returns channel attribution based on the specified model.

Description: "Returns channel attribution analysis (first touch, last touch, linear, or time decay) for the specified time period, including conversions, revenue, and ROI by channel."

SQL Query:

SELECT 
  CASE 
    WHEN {model} = 'first_touch' THEN first_touch_channel_calc
    WHEN {model} = 'last_touch' THEN last_touch_channel_calc
    ELSE 'multi_touch'
  END as attributed_channel,
  COUNT(DISTINCT contact_id) as conversions,
  SUM(deal_amount) as total_revenue,
  SUM(total_marketing_spend) as total_spend,
  AVG(roi_percent) as avg_roi
FROM marketing_attribution_view
WHERE conversion_date >= DATE_SUB(CURRENT_DATE(), INTERVAL {period} DAY)
GROUP BY attributed_channel
ORDER BY conversions DESC;

Tool 2: get_customer_journey(contact_id: string)

This tool returns the complete customer journey for a specific contact.

Description: "Returns complete customer journey including all touchpoints, channels, campaigns, and conversion path for a specific contact."

SQL Query:

SELECT * FROM marketing_attribution_view
WHERE contact_id = {contact_id}
ORDER BY touchpoint_date ASC;

Tool 3: analyze_channel_roi_with_usage(channel: string, period: string)

This tool analyzes channel ROI including product usage value.

Description: "Returns ROI analysis for a specific channel including product usage value, power user rates, and lifetime value calculations."

SQL Query:

SELECT 
  channel,
  COUNT(DISTINCT contact_id) as conversions,
  SUM(total_marketing_spend) as total_spend,
  SUM(deal_amount) as total_revenue,
  AVG(roi_percent) as avg_roi,
  COUNT(DISTINCT CASE WHEN user_segment = 'power_user' THEN contact_id END) as power_users,
  COUNT(DISTINCT CASE WHEN user_segment = 'power_user' THEN contact_id END) * 100.0 / COUNT(DISTINCT contact_id) as power_user_rate,
  AVG(roi_with_usage_value) as avg_roi_with_usage
FROM marketing_attribution_view
WHERE first_touch_channel_calc = {channel}
  OR last_touch_channel_calc = {channel}
  AND conversion_date >= DATE_SUB(CURRENT_DATE(), INTERVAL {period} DAY)
GROUP BY channel;

Tools can be called by agents with natural language

Step 5: Wire the Tool to an Agent Builder

Connect your Pylar tools to an agent builder. You can use:

• OpenAI Platform – Build custom agents with GPT-4
• LangGraph – Build complex agent workflows
• Claude Desktop – Quick setup for internal use
• Cursor – Developer-focused agent builder
• Zapier / Make / n8n – Automation platforms

Agent Instructions:

Give your agent clear instructions.

You are a marketing attribution assistant. Your job is to help marketing teams understand channel performance, customer journeys, and attribution across their marketing stack.

Rules:

1. Only use the tools provided. Do not invent attribution data or make assumptions.
2. When asked about channel attribution, use get_channel_attribution with the requested model and time period.
3. When asked about a specific customer journey, use get_customer_journey to get complete context.
4. When asked about channel ROI, use analyze_channel_roi_with_usage to include product usage value.
5. Always provide context: explain which channels drive conversions, what the customer journey looks like, or how ROI changes when you include usage data.
6. If you don't have data, say so. Don't guess.

Available tools:

• get_channel_attribution(model, period) — Get channel attribution analysis
• get_customer_journey(contact_id) — Get complete customer journey
• analyze_channel_roi_with_usage(channel, period) — Analyze channel ROI with usage data

Example Prompts Users Can Try:

1. "Which channels drove the most conversions this quarter?"
2. "What's the customer journey for customers who converted from our Q1 campaign?"
3. "How do first-touch and last-touch attribution compare for our email campaigns?"
4. "Which marketing channels have the best ROI when you include product usage data?"
5. "What's the attribution breakdown for customers who became power users?"
6. "Which campaigns are driving customers who actually use the product?"

The agent uses the tools to answer these questions, providing attribution analysis and actionable insights.

Step 6: Add Evals & Guardrails

Add observability and guardrails to ensure the agent works correctly and safely.

Logging:

Track:

• Which channels and campaigns are queried most frequently
• Error rates and types
• Query latency
• Unusual outputs or patterns

Use Pylar's Evals system to monitor:

• Success rates: How often queries succeed vs. fail
• Error patterns: What errors occur and why
• Query shapes: What types of queries agents make
• Raw logs: Every query with full context

Guardrails:

Add simple checks:

No PII leaking: Views exclude sensitive data (personal email addresses in raw form, internal campaign notes). Agents can only query through views, so they never see PII.

Row limits: Views automatically limit results (e.g., top 100 channels, last 90 days of conversions). Agents can't query unbounded datasets.

Time windows: Views filter to recent data (last 90 days for conversions, last 30 days for usage). Agents can't query historical data beyond what's needed.

Access control: Different views for different teams. Marketing teams see all attribution data. Finance sees only aggregated ROI. Product sees only usage-related attribution.

Step 7: Rollout Playbook

How to introduce this agent to your marketing team:

Week 1: Pilot with 2–3 marketers

• Set up the agent with basic tools
• Have marketers try it for channel analysis
• Collect feedback on what's missing

Week 2: Expand to full team

• Add more tools based on feedback
• Create team-wide Slack integration
• Share example prompts

Week 3: Add leadership

• Create executive views (aggregated ROI, channel mix)
• Add leadership tools (budget allocation, channel optimization)
• Set up weekly attribution reviews using the agent

Example Slack message: "Before your next channel review, ask the attribution agent: 'Which channels drove the most conversions this quarter and what's their ROI?'"

How This Replaces Dashboards (Without Killing Them)

Old flow: Export data from Google Ads, Facebook, email platform, CRM, product analytics. Import into spreadsheet. Manually match conversions to channels. Calculate attribution. Build charts. Spend 4 hours to answer one question.

New flow: Ask one question, get complete attribution analysis with channel performance, customer journeys, ROI calculations, and usage insights—all in one answer.

The agent synthesizes data from multiple systems into a coherent narrative. You get answers in seconds instead of hours.

Dashboards are still useful for:

• Monitoring known metrics
• Executive summaries
• Scheduled reports
• Visual exploration

Agents are better for:

• Ad-hoc questions
• Cross-system analysis
• Deep dives
• Proactive insights

Use both. Dashboards for monitoring. Agents for exploration.

Most marketing teams struggle to understand which channels actually drive conversions. This agent automatically tracks customer journeys, attributes conversions to channels, and surfaces insights across your entire marketing stack.

If you want to try this with your own data, you can spin up a Marketing Attribution View in Pylar and plug it into your existing agent builder. Start with CRM and one marketing platform, add more channels and product usage data as you connect them, and iterate based on what your team needs.