---
title: "Churn prediction for B2B SaaS: what works and how to test it"
url: https://renewalfix.com/guides/churn-prediction-b2b-saas/
description: "How churn prediction works in B2B SaaS: health score rules, machine learning and AI that reads conversations. The data you need and how to test it."
updated: 2026-11-03
---

Guide

# Churn prediction for B2B SaaS: what works and how to test it

Churn prediction is the practice of flagging the accounts likely to cancel or shrink at renewal, early enough to act. In B2B SaaS it works differently from consumer churn prediction: fewer customers, bigger accounts, and most of the warning in conversations rather than clicks.

By [John Beckett](https://renewalfix.com/about/#john-beckett) · Updated November 3, 2026

## Three approaches

| Approach | How it works | Strong when | Watch out for |
| --- | --- | --- | --- |
| Rules and health scores | Points or bands from usage, tickets, [NPS](https://renewalfix.com/glossary/net-promoter-score/) and CRM fields | You want a score everyone can understand and audit | Averages that let one weak area hide; stale manual fields |
| Machine learning on outcomes | A model learns which patterns came before past churn | You have many renewals and clean history | Too few churn events to learn from; hard to explain a single score |
| AI that reads conversations | Language models read calls, tickets and email for risk | Most of the warning is in what customers say | Made-up or unsupported claims unless every risk must cite its source |

They aren't exclusive. A practical setup uses AI to find evidence in conversations, and plain rules to turn that evidence into a score people can check.

## The data you need

- **CRM:** accounts, [ARR](https://renewalfix.com/glossary/annual-recurring-revenue/), renewal dates, owners, and closed renewals (won and lost) to test against.
- **Conversations:** call recordings, support tickets, email.
- **Product:** seat and feature usage over time.
- **Engineering:** bugs and blockers linked to accounts.
- **Billing:** payment history, late invoices and requested downgrades.

## How to test any churn prediction

Run it on last year's renewals as they looked before the renewal date, then check:

1. **Coverage of churned revenue (recall):** what share of the ARR that actually churned was flagged 90, 60 and 30 days before the renewal?
2. **Precision:** of the accounts flagged Red, how many actually churned or shrank?
3. **Lift:** how many times more likely is a Red account to churn than the average account?
4. **Lead time:** how early did the first warning appear?

Judge precision against your base churn rate. If 10% of your renewals churn and 30% of the accounts flagged Red churned, the score finds churn 3 times better than picking accounts at random.

**Avoid leakage** (using information the model wouldn't have had at the time). If the model uses a cancellation notice to "predict" a cancellation, it will look accurate and help no one. Test with what was known before the decision.

## Questions to ask a vendor

- Can a [CSM](https://renewalfix.com/glossary/customer-success-manager/) see why an account is at risk, down to the call, ticket or email?
- How do you treat accounts with little data?
- Can we backtest on our own history before we switch?
- How will we know the actions we take actually changed the outcome?

Horizon combines AI that reads conversations with a rules-based [health score](https://renewalfix.com/product/customer-health-score/), lets admins backtest it against past churn, and measures actions against a [holdout group](https://renewalfix.com/product/holdout-measurement/).

More on this topic: [churn, definitions, guides and tools](https://renewalfix.com/topics/churn/).

---

Horizon is renewal risk software for Customer Success teams. Book a call: https://calendar.chilipiper.com/book/me/serban-goanta
