All case studies
Customer Success·Illustrative Client Scenario·SaaS · Series C, ~$60M ARR

Predictive health scoring for a Series C SaaS platform

Replacing a colour-coded spreadsheet with a usage-, sentiment-, and outcome-based health model — so CSMs could intervene 30+ days earlier with the right play.

30+ days
earlier identification of at-risk accounts
+4.2 pts
gross retention improvement (illustrative)
60%
of CSM week reallocated from reporting to action
Challenge

The situation

  • Gross retention had slipped from 92% to 86% over four quarters and renewals were being surprised in the final 60 days.
  • Health scores were maintained manually by CSMs in a spreadsheet and rarely matched what later happened at renewal.
  • Product usage, support sentiment, and outcome data lived in three different systems with no shared definition of 'healthy.'
Approach

How we engaged

01

Diagnose

Interviewed 18 CSMs and 12 renewed/churned customers to map the real leading indicators of churn — not the ones in the existing rubric.

02

Design

Built a weighted health model across product engagement, outcome attainment, executive engagement, and support sentiment, with explicit thresholds and recommended plays per segment.

03

Deliver

Implemented the model in the CS platform, instrumented the missing product events, and wired health-triggered tasks into the CSM workflow.

04

Operate

Established a weekly health-review cadence and a quarterly model-tuning ritual co-owned by CS Ops and Product Analytics.

Solution

What we built and ran

  • A four-pillar health model (Adoption · Outcomes · Relationship · Sentiment) with segment-specific weights
  • An at-risk playbook library mapped to the dominant failure modes per segment
  • Health-triggered task automation embedded into the CSM daily workflow
  • An executive scorecard with cohort-level health trend, save rate, and forecast retention
Lessons learned

What we'd carry forward

  • Health scoring fails when it tries to be a single number. Pillars + thresholds drive better behaviour than weighted averages.
  • If a CSM can't act on a score by tomorrow morning, the score is the wrong unit of work — design plays, not dashboards.
  • Tune the model quarterly. Drift is silent and erodes trust faster than missing the next renewal.

Want a tailored example for your context?

Tell us about your industry and the outcomes you're after. We'll send a relevant scenario and a recommended path forward.