Use case · C-Risk
Find where churn is born — before it happens.
Nearly every customer who churned showed up in your feedback first — the risk was there, it just wasn’t scored. The complaint was logged, categorised, maybe even resolved. But the risk it signalled — the likelihood that this specific complaint would escalate into churn, public criticism, or legal action — was never assessed.
Complaint-Risk Scoring (C-Risk) changes that. Every piece of feedback is scored on a two-dimensional risk matrix:
Likelihood of escalation
Will this customer complain louder, leave, or take action?
Severity of impact
How damaging is the underlying issue to the business?
The result is a prioritised queue. Not sorted by complaint volume. Sorted by risk.
Risk matrix
Likelihood of escalation × severity of impact
“I’ve emailed 3 times about this issue”
Risk 0.61
“This is a data breach and I’m contacting authorities”
Risk 0.94
“Shipping was 2 days late”
Risk 0.15
“Overcharged by $50”
Risk 0.42
The volume trap
Why sorting complaints by volume is dangerous
Most CX teams prioritise by volume. “We got 200 complaints about shipping this month — that’s our #1 priority.” But volume doesn’t equal risk.
Consider: 200 shipping complaints, mostly minor delays, sentiment mildly negative, no churn signals. Low risk. Meanwhile, 8 complaints mention billing errors, legal language, and explicit threats to contact regulators. High risk — but invisible in a volume-sorted queue.
C-Risk flips the priority. It scores each complaint individually, then aggregates risk by theme. You see not just what customers complain about most, but which complaints have the highest potential to damage the business.
| Volume-sorted | C-Risk sorted | |
|---|---|---|
| #1 priority | Shipping (200 complaints, low severity) | Billing errors (8 complaints, critical severity) |
| Action | Fix shipping delays | Legal review + immediate resolution |
| Outcome | Marginal improvement | Prevented regulatory complaint |
The risk matrix
How C-Risk scores complaints
Every complaint receives two scores.
Likelihood of escalation (0–1.0)
Based on linguistic signals in the feedback: urgency language, repetition, threat indicators, mentions of competitors, escalation language (“I’ve contacted you three times,” “This is my last email,” “I’m filing a complaint with…”).
Severity of impact (0–1.0)
Based on the nature of the issue: financial impact, data/privacy concerns, safety/health, legal exposure, brand/reputation risk. Weighted by customer segment value.
“I’ve emailed 3 times about this issue”
Risk 0.61
High likelihood, low severity. The customer is frustrated but the issue is minor. A quick response prevents escalation.
“This is a data breach and I’m contacting authorities”
Risk 0.94
High likelihood, high severity. Immediate action required. Route to senior team, legal, or executive intervention.
“Shipping was 2 days late”
Risk 0.15
Low likelihood, low severity. Track for pattern changes. No action needed.
“Overcharged by $50”
Risk 0.42
Low likelihood, high severity. The issue is serious but the customer isn’t escalating. Proactive resolution prevents future problems.
Upload tickets, reviews, or call transcripts — Scoty runs C-Risk automatically.
Worked example
An illustrative SaaS company analysing 3,200 support tickets
Mid-market B2B SaaS company. 3,200 support tickets over 90 days. 280 contained complaint language. C-Risk was run on all 280.
| Rank | C-Risk | Theme | Sample language | Action taken |
|---|---|---|---|---|
| 1 | 0.94 | Data Privacy | “considering legal action re: data” | Escalated to legal + DPO |
| 2 | 0.89 | Billing Dispute | “filing chargeback with bank” | Finance + CS joint resolution |
| 3 | 0.87 | Service Outage | “lost a client because of downtime” | Executive outreach, SLA credit |
| 4 | 0.82 | Security Concern | “reporting to ICO” | Security team investigation |
| 5 | 0.78 | Contract Terms | “my lawyer reviewed the T&Cs” | Legal review of terms |
| 6 | 0.74 | Performance | “switching to competitor next week” | Retention team intervention |
| 7 | 0.71 | Data Loss | “lost 3 hours of work, no recovery” | Engineering hotfix + apology |
| 8 | 0.69 | Accessibility | “unable to use the product, ADA” | Product accessibility audit |
| 9 | 0.66 | Integration Failure | “costing us roughly $2,000/day in lost revenue” | Priority engineering ticket |
| 10 | 0.64 | Support Quality | “worst support experience, leaving” | CS Director personal follow-up |
What happened
Complaints ranked 1–5 were previously invisible — buried in the support queue with no special routing. The data privacy complaint (#1) had been open for 11 days. The billing dispute (#2) had been auto-closed as “resolved.”
After C-Risk: all 10 were escalated within 24 hours. Complaints 1, 2 and 4 were resolved without escalation to external authorities. Complaint 6 (churn risk) was acted on within 24 hours via retention intervention.
The volume-sorted view would have ranked “Performance” and “Support Quality” as top priorities (highest complaint counts). The five highest-risk complaints would have remained in the general queue.
Comparison
C-Risk vs. traditional churn models
| Traditional churn model | C-Risk |
|---|---|
| Based on behavioural data (usage, login frequency) | Based on feedback content (what customers say) |
| Predicts likelihood of churn | Predicts likelihood of escalation (churn, public complaint, legal) |
| Requires historical churn data to train | No training data needed — runs on any complaint dataset |
| Monthly or quarterly prediction cycle | Real-time — scores every complaint as it’s analysed |
| Doesn’t explain why a customer will churn | The verbatim explains the risk — you know the trigger |
| Segment-level prediction | Individual complaint-level scoring |
| Needs a data science team to maintain | No data science required — automated pipeline |
C-Risk doesn’t replace behavioural churn models. It complements them. Behavioural models tell you who might leave. C-Risk tells you why and how urgently.
Better together
The full picture: ABSA × C-Risk
Run ABSA and C-Risk together and you get a matrix that connects experience quality to business risk.
| Aspect | Sentiment | High-risk complaints | Action |
|---|---|---|---|
| Billing | −0.35 | 12 (0.70+ risk) | Immediate — billing system audit |
| Data Privacy | −0.28 | 4 (0.85+ risk) | Critical — legal & security review |
| Support Quality | −0.42 | 8 (0.60+ risk) | Urgent — staffing & training |
| Performance | −0.15 | 3 (0.65+ risk) | Monitor — engineering review |
| Product Quality | +0.62 | 1 (0.55 risk) | None — protect this |
This is what your executive dashboard should show: not just sentiment, not just risk — both, overlaid. It tells you exactly where to spend your next sprint.
Every day without C-Risk, high-risk complaints sit in your queue. Unseen.
Run C-Risk on your last 90 days of support tickets and see what’s been hiding in plain sight.
Runs on tickets, reviews, and call transcripts — hosted in the EU or Switzerland (Infomaniak) when you choose. Every risk score links back to the verbatim behind it.
EU/Swiss hostingGDPR-readyEvidence-linked scores