From CRM Data to Decisions: How to Build a Simple Churn Detection System in HubSpot

Most teams believe churn prediction requires machine learning. It does not. The signals are already in HubSpot — what is missing is structure. This article describes the three-component system that turns raw HubSpot data into churn decisions, with a working example you can implement this week.

Quick answers

How do you build a churn detection system in HubSpot?: Build a HubSpot churn detection system with three components: (1) Signals — raw data points like last_reply_date and contract_end_date, (2) Conditions — thresholds that turn signals into meaning (no reply for 30 days), and (3) Decisions — outputs like creating a HubSpot task for the company owner. Run the system on a recurring schedule against every company record.
Do you need machine learning to predict churn in HubSpot?: No. Machine learning is not required to predict churn in HubSpot. A rule-based system using existing CRM fields (renewal date, last reply, meeting count) detects 80–90% of churn risk with zero training data. Rule-based systems also produce explainable evidence, which ML models do not.
What HubSpot data is needed for churn detection?: Four HubSpot fields are sufficient for churn detection: contract_end_date on the company or deal, last inbound email reply date, count of upcoming and completed meetings (last 30 days), and HubSpot Company Owner. No custom properties required for a baseline system.
What is the simplest churn detection rule for HubSpot?: The simplest effective HubSpot churn detection rule: IF contract_end_date is within 90 days AND upcoming_meetings = 0 AND last_inbound_reply is older than 30 days THEN flag the account as at risk and create a task for the company owner with the evidence in the body.
How often should a churn detection system run?: A churn detection system should run at least every 24 hours, ideally every 60 minutes. Hourly evaluation catches new risk windows as soon as the renewal date crosses the 90-day boundary, which is the typical intervention window for B2B SaaS.

What is a churn detection system?

A churn detection system is a recurring evaluation of CRM data that classifies each customer account as either at risk or not at risk and produces a concrete next action. In HubSpot it consists of three components: signals (raw data), conditions (thresholds), and decisions (outputs).

The three components

1. Signal

The raw HubSpot data point: last_reply_date, meeting count, deal stage, contract_end_date, last_activity_date.

2. Condition

The threshold that turns a signal into meaning: "no reply for 30 days," "no meeting scheduled within the renewal window."

3. Decision

The output: flag the account as at risk and create a HubSpot task for the company owner with the triggering evidence in the body.

Example system: renewal risk

A working example for renewal risk:

Scope: all HubSpot companies with contract_end_date in the next 90 days.
Condition: upcoming_meetings = 0 AND last_inbound_reply > 30 days ago.
Decision: create a HubSpot task assigned to the Company Owner with body: "Renewal in N days · 0 meetings scheduled · last reply X days ago."
Auto-close: when a meeting is scheduled or an inbound reply arrives, close the task automatically.

This is the entire system. Three components, four fields, one task per matched account.

Useful beats perfect

The biggest mistake teams make is building perfect models instead of useful systems. A basic system that runs every 60 minutes and creates the right task is more valuable than an unused ML model that produces an unexplained score.

Sighub is built exactly on this principle: signals, conditions, decisions, automated across every company record in your HubSpot portal. Read next: 5 HubSpot Signals That Predict Churn or Why HubSpot Tasks Are the Missing Execution Layer.