Designing a Subscription Churn Classification System

A SaaS company currently counts only explicit cancellations as churn — reporting a 2% monthly churn rate. But the analyst knows there are at least 3 other churn patterns (payment failure, silent inactivity, and downgrade). She must design a rule-based churn classification system and manually trace through 12 subscriber histories before the board presentation.

# Aplly.xyz Case Study Submission

## Title
Designing a Subscription Churn Classification System

## Type
Data Analytics

## Difficulty
Intermediate

## Estimated Time
60 minutes

## Overview
A SaaS company currently counts only explicit cancellations as churn — reporting a 2% monthly churn rate. But the analyst knows there are at least 3 other churn patterns (payment failure, silent inactivity, and downgrade). She must design a rule-based churn classification system and manually trace through 12 subscriber histories before the board presentation.

## Case Details

Function Focus: Rule-based churn classification design, multi-type churn taxonomy, borderline-case detection, churn rate calculation under multiple definitions

Scenario:
SaaSify's board presentation claims "monthly churn is only 2%." This number counts only voluntary cancellations. The analyst knows the real picture is more complex: some subscribers stop paying (payment failure churn), some keep paying but never log in (silent churn), and others downgrade to a free plan (revenue churn). She needs a classification system with clear rules for each churn type, then must manually trace 12 subscriber histories to test whether the rules capture every churn pattern or miss important edge cases.

Churn Categories (Draft):
1. Voluntary Churn — subscriber explicitly cancels subscription via any channel
2. Payment Failure Churn — subscription auto-cancelled after 3+ failed payment retries
3. Silent Churn — subscription paid but no login activity for 90 consecutive days
4. At Risk — subscriber is still active but shows warning signs (declining login frequency, support tickets expressing dissatisfaction)
5. Active — none of the above

Draft Classification Rules:
- Rule 1: If cancellation event exists → Voluntary Churn (regardless of other signals)
- Rule 2: If no cancellation but account status = suspended due to payment → Payment Failure Churn
- Rule 3: If subscription active and paid, but days since last login ≥ 90 → Silent Churn
- Rule 4: If days since last login ≥ 60 but < 90, OR support tickets mentioning "price" or "cancel" in last 30 days → At Risk
- Rule 5: Everything else → Active
- Priority order: Rule 1 > Rule 2 > Rule 3 > Rule 4 > Rule 5

Tasks:
1. Before tracing, read the draft categories and rules — identify at least 2 gaps (e.g., what about subscribers who cancel but keep using the free tier? what about subscribers who downgrade?)
2. Manually classify all 12 subscriber histories using the draft rules — produce a table with subscriber ID, assigned category, rule that matched, and any uncertainty
3. Identify 3 subscribers that are difficult to classify under the draft rules — explain why they don't fit cleanly and propose a rule revision for each
4. Using your classifications, compute the churn rate under two definitions: (a) narrow churn = Voluntary + Payment Failure only, and (b) broad churn = Voluntary + Payment Failure + Silent Churn — explain why the board's 2% number is misleading and what the real range is
5. Only after submitting your manual analysis, compare your classification against a tool-based implementation (spreadsheet logic or AI classifier) and report any disagreements — finalize your recommended rule set

Expected Output:
A churn classification design packet containing: (a) critique of the draft taxonomy with at least 2 gaps identified, (b) 12-subscriber classification table with rule references, (c) 3 borderline cases with rule revision proposals, (d) narrow vs broad churn rate comparison showing why the board's number is misleading, and (e) a final recommended rule set with round-trip discrepancy note.

Evaluation Criteria:
Completeness of the 12-subscriber classification, correct identification of gaps in the draft taxonomy, quality of rule revision proposals for borderline cases, correct churn rate arithmetic under both definitions, and a clear argument about why the narrow churn rate is misleading.

## Data Sources

Subscriber Histories (12):
| ID | Plan | Days Since Signup | Last Login (day) | Status | Cancellation Event | Failed Payments | Support Tickets (last 60 days) |
|---|---|---|---|---|---|---|---|
| S01 | Premium | 210 | 210 | Active | Cancelled day 180 ("too expensive") | 0 | 2 ("how do I cancel?") |
| S02 | Basic | 240 | 200 | Suspended | None | 3 failed, suspended day 210 | 0 |
| S03 | Premium | 200 | 45 | Active | None | 0 | 0 |
| S04 | Pro | 180 | 178 | Active | None | 0 | 1 ("feature request") |
| S05 | Basic | 180 | 90, then 92, then 180 | Active | None | 0 | 0 |
| S06 | Premium | 150 | 148 | Active | None | 0 | 4 ("considering alternatives") |
| S07 | Basic | 120 | 118 | Active | None | 1 (retried next day, succeeded) | 0 |
| S08 | Premium → Basic | 200 | 190 | Active | None (downgraded day 150) | 0 | 0 |
| S09 | Free Trial | 30 | 28 | Expired | Cancelled after trial | 0 | 0 |
| S10 | Basic | 200 | 10 | Cancelled | Auto-cancelled day 190 (inactivity per T&C) | 0 | 0 |
| S11 | Premium | 180 | 140 | Active | None | 0 | 0 |
| S12 | Pro | 365 | 360 | Deleted | Account deleted day 360 by user request | 0 | 1 ("please delete my data") |

Context:
- Board reported churn rate: 2% (defined as voluntary cancellations only)
- Monthly active subscribers: 5,000
- Average revenue per subscriber: ₹2,400/month
- Free trial converts to paid at day 31 if payment method added

Key Dates:
- Current day: day 365 of operations
- Silent churn threshold: 90 days (industry standard for monthly SaaS)
- Payment retry policy: 3 attempts over 7 days, then suspend
- Inactivity auto-cancel policy: 180 days without login → auto-cancel

## Solution Frameworks
Rule-based classification system, churn taxonomy design, borderline-case analysis, churn rate calculation under multiple definitions, metric communication

## Solver Guidance & Tutorials
Link to: "Beyond Cancellations — Building a Multi-Type Churn Classification" tutorial

## What You'll Learn
- Designing a churn taxonomy that captures voluntary, passive, and payment-failure churn
- Detecting borderline cases that don't fit simple rules
- Computing churn rates under multiple definitions
- Communicating why a narrow churn definition misleads leadership

## Tags
churn classification, subscription analytics, customer retention, rule-based system, SaaS metrics

## Registration Links
- Register as Solver
- Register as Evaluator

Subscriber Histories (12):
| ID | Plan | Days Since Signup | Last Login (day) | Status | Cancellation Event | Failed Payments | Support Tickets (last 60 days) |
|---|---|---|---|---|---|---|---|
| S01 | Premium | 210 | 210 | Active | Cancelled day 180 ("too expensive") | 0 | 2 ("how do I cancel?") |
| S02 | Basic | 240 | 200 | Suspended | None | 3 failed, suspended day 210 | 0 |
| S03 | Premium | 200 | 45 | Active | None | 0 | 0 |
| S04 | Pro | 180 | 178 | Active | None | 0 | 1 ("feature request") |
| S05 | Basic | 180 | 90, then 92, then 180 | Active | None | 0 | 0 |
| S06 | Premium | 150 | 148 | Active | None | 0 | 4 ("considering alternatives") |
| S07 | Basic | 120 | 118 | Active | None | 1 (retried next day, succeeded) | 0 |
| S08 | Premium → Basic | 200 | 190 | Active | None (downgraded day 150) | 0 | 0 |
| S09 | Free Trial | 30 | 28 | Expired | Cancelled after trial | 0 | 0 |
| S10 | Basic | 200 | 10 | Cancelled | Auto-cancelled day 190 (inactivity per T&C) | 0 | 0 |
| S11 | Premium | 180 | 140 | Active | None | 0 | 0 |
| S12 | Pro | 365 | 360 | Deleted | Account deleted day 360 by user request | 0 | 1 ("please delete my data") |

Context:
- Board reported churn rate: 2% (defined as voluntary cancellations only)
- Monthly active subscribers: 5,000
- Average revenue per subscriber: ₹2,400/month
- Free trial converts to paid at day 31 if payment method added

Key Dates:
- Current day: day 365 of operations
- Silent churn threshold: 90 days (industry standard for monthly SaaS)
- Payment retry policy: 3 attempts over 7 days, then suspend
- Inactivity auto-cancel policy: 180 days without login → auto-cancel

Rule-based classification system, churn taxonomy design, borderline-case analysis, churn rate calculation under multiple definitions, metric communication