Identifying the behavioral and operational signals most closely associated with churn in a fictional B2B SaaS customer base.
MorpheusGroup has 500 SaaS customers. The goal was to identify which signals could help a Customer Success team prioritize churn risk earlier and respond with the right intervention.
I treated this as a Customer Success prioritization problem rather than simply looking for the highest churn percentage. The aim was to understand which signals were useful, how consistently they separated churned from retained customers, and whether combining them produced a more actionable risk profile.
I first checked churn by plan and broad account characteristics to understand the dataset before deciding where to dig deeper.
I grouped feature adoption and seat utilization into consistent ranges, then calculated the churn rate within each cohort.
I grouped accounts by unresolved-ticket count to see whether accumulating support issues corresponded with materially higher churn.
I then crossed adoption, utilization, and unresolved-ticket levels to identify risk profiles that would be useful for CSM prioritization.
I also tested tenure, renewal timing, sentiment/CSAT, and other dimensions. I kept them out of the final risk story when they did not show useful separation.
Finally, I ranked the strongest signals by practical value and mapped each one to a Customer Success intervention rather than stopping at the numbers.
CASE
WHEN feature_adoption < 40 THEN 'Under 40%'
WHEN feature_adoption < 60 THEN '40–59%'
WHEN feature_adoption < 80 THEN '60–79%'
ELSE '80%+'
END AS feature_adoption_band
GROUP BY feature_adoption_band
ORDER BY churn_rate_pct DESC;The same approach was repeated for seat utilization and unresolved support issues before the variables were combined into multi-signal risk segments.
The strongest operational warning signal across the analysis.
The strongest product-engagement signal, with churn falling consistently as adoption increased.
A strong supporting engagement signal, especially when combined with unresolved support issues.
| Feature adoption | Churn rate |
|---|---|
| Under 40% | 43.8% |
| 40–59% | 28.5% |
| 60–79% | 15.8% |
| 80%+ | 11.6% |
| Seat utilization | Churn rate |
|---|---|
| Under 40% | 41.1% |
| 40–59% | 21.8% |
| 60–79% | 16.3% |
| 80%+ | 12.0% |
| Unresolved tickets | Customers | Churn rate |
|---|---|---|
| 0 | 310 | 17.1% |
| 1 | 129 | 19.4% |
| 2 | 51 | 31.4% |
| 3+ | 10 | 80.0% |
The 3+ group is small, so the 80% result should be interpreted cautiously. The broader pattern still points to unresolved issues as an important churn-risk signal.
This segment contained 23 customers and was the strongest meaningful combined-risk profile found in the analysis.
Little separation in churn across tenure groups.
No consistent progression as renewal approached.
Comparatively weak or noisy as standalone indicators.
Multiple unresolved tickets should trigger immediate support escalation and proactive CSM follow-up.
Use-case reviews, enablement, and targeted training for low-adoption accounts.
Check whether low usage reflects adoption gaps, over-provisioning, or weaker product fit.
Some highly engaged customers still churned. To investigate those cases, I would add structured cancellation reasons, pricing feedback, competitor data, stakeholder changes, budget constraints, and contract history.
The analysis was completed in SQL using a synthetic 500-account dataset. I tested plan-level churn, product engagement, support friction, CSM contact, tenure, renewal timing, and combined-risk segments before deciding which signals were useful enough to keep.