Predictive churn → marketing save (L5)

The steps

1. 01

   Architect the unified health score

   The RevOps Manager or Data Analyst (Owner) must build a unified data view in a warehouse like Snowflake or BigQuery. A 'maturity level 5' health score cannot rely on a single metric; it requires three pillars: Product Usage (via Segment or Mixpanel), Support Sentiment (via Zendesk/Gorgias), and Billing Health (via Stripe/NetSuite). • Actions: Join table 'subscriptions' with 'product_events' (log-ins, feature usage) and 'support_tickets' (volume and sentiment). • Logic: Create a weighted calculation. Example: Usage Frequency (40%) + Ticket Urgency/CSAT (30%) + Overdue Invoices (30%). • Setup: Use a tool like Census or Hightouch to sync these composite scores back into your CRM (Salesforce/HubSpot) under a custom field called 'Predictive_Health_Score_Current'. • Pitfall: Avoid 'vanity' metrics like NPS; customers often provide high NPS right before churning due to budget cuts. • Time Estimate: 10-15 hours. • Definition of Done: A live dashboard showing a 0-100 score for every active account updated every 4 hours.

2. 02

   Configure the volatility trigger

   The GTM Ops specialist must configure an automation trigger that monitors for volatility, not just low scores. Churn prediction is most accurate when it catches a 'rapid decline.' • Logic: Set a trigger when ('Health_Score_Current' < 'Health_Score_7_Day_Avg' - 15). This captures a 15-point drop in a 24-hour window. • Configuration: In HubSpot, use a 'Workflow' triggered by 'Property Value Change'. In Salesforce, use 'Flow Builder' with an 'Entry Condition' monitoring the health score field. • Action: When the drop occurs, the automation should immediately add the contact to a 'Marketing At-Risk Segment' and notify the assigned CSM via Slack. • Pitfall: Don't trigger on accounts already in 'Onboarding' status, as scores fluctuate naturally during setup. • Time Estimate: 3-5 hours. • Definition of Done: A test record with a simulated 20-point drop successfully triggers a Slack alert and a CRM segment membership change within 15 minutes.

3. 03

   Design multi-channel save plays

   The Growth Marketer (Owner) develops three distinct pathways for the 'Save Play' based on the reason for the score drop. • Pathway A (Usage Drop): Automated email from the Founder/CEO offering a functional training session. • Pathway B (Support Friction): A 'Concierge' ticket creation that bypasses the standard queue. • Pathway C (Billing Issues): A 'Grace Period' extension offer sent via SMS or Email. • Specifics: Use a tool like Mutiny or Jasper to create variations of these messages. • Example Prompt for AI: 'Write a high-empathy email to a B2B SaaS user who hasn't logged in for 10 days. Offer a 15-minute optimization call. Sound helpful, not stalker-ish.' • Pitfall: Sending an 'automated-looking' email. Use plain-text formatting to ensure it looks like a 1-to-1 message from an executive. • Time Estimate: 8 hours. • Definition of Done: Three distinct sequences live in the Marketing Automation Platform (MAP) ready to receive at-risk leads.

4. 04

   Establish the control group environment

   To prove the Save Play works, the Data Analyst must implement a 'Holdback Control' group. • Action: In your automation tool (e.g., Braze or HubSpot), use a 'Random Split' node. Allocate 90% of at-risk accounts to the 'Save Play' (Treatment) and 10% to a 'Control' group that receives no special marketing intervention. • Measurement: Create a report comparing the 'Account Retention Rate' between these two groups after 30 days. • Configuration: Ensure the 'Control' group flag is stamped on the CRM record to prevent manual CSM intervention from skewing the marketing experiment. • Pitfall: Skipping the control group because it feels 'wrong' to let 10% churn; without it, you cannot calculate the incremental ROI of your marketing spend. • Time Estimate: 2-4 hours. • Definition of Done: A reporting dashboard showing the 'Lift' (Treatment Retention vs. Control Retention) in real-time.

5. 05

   Tie outcomes to NRR and ROI

   Finally, the RevOps Lead must tie the success of these plays to Net Revenue Retention (NRR) rather than 'Saved Accounts.' • Action: Pull the MRR (Monthly Recurring Revenue) value of the accounts entering the Save Play. Calculate the 'Saved Dollar Value' = (MRR of Treatment Group that didn't churn) minus (MRR of Control Group that would have churned anyway). • Calculation: Formula: (Successful Saves x Avg Contract Value) / (Cost of Save Play Incentives + Labor). • Goal: The Save Play should contribute to a 2-5% increase in total NRR. • Pitfall: Over-incentivizing with discounts. If you save an account by giving a 50% discount, your NRR might still suffer. Prioritize 'Value Adds' (e.g., extra seats, premium support) over 'Price Cuts.' • Time Estimate: 5 hours (setup for ongoing reporting). • Definition of Done: Monthly board-level report showing exactly how many NRR percentage points are attributable to the AI-triggered Save Play.

Next playbooks

• AI customer-interview synthesis (L3)
• AI SEO clusters with Clearscope/Frase (L3)
• AI-generated event follow-up (L3)
• Lifecycle email AI personalization (L3)