Project Overview
Churn isn’t equal across customers, and neither is its financial impact. Most retention strategies ignore this and spread effort too thin.
This project takes a different approach. It focuses on identifying high-impact customer segments, understanding why they leave, and prioritizing actions that deliver measurable returns. The goal is simple: reduce churn where it matters most, and do it in a way that makes economic sense.
The Approach and Process
Methodology
The analysis is built on a simple idea: not all churn matters equally, and not all churn is preventable. The work moves through four stages.
- First, a clean and consistent customer dataset is built. It brings together tenure, contract type, service usage, payment behavior, churn status, and lifetime value. This creates a solid base for segmentation.
- Second, the analysis looks across key behavioral dimensions, contract type, internet service, and payment method, to uncover patterns behind churn.
- Third, a financial layer is added by calculating value at risk(VaR). This combines churn probability with customer lifetime value to estimate potential revenue loss.
- Finally, customers are grouped into priority segments, each tied to specific retention actions and expected ROI.
This structure keeps the strategy grounded in three things: customer behavior, financial impact, and execution clarity.
Analysis & Findings by Stage
Stage 1 — The Churn Timing: Understanding When Customers Leave
Problem Statement: The first step answers a simple question: when are customers most likely to churn? Timing matters because retention works best before customers fully decide to leave.
Analytical Approach: Customers are grouped into tenure bands (0–3 months, 4–6 months, up to 36+ months). This reflects natural lifecycle stages such as onboarding, early usage, and long-term retention.
Key Findings: Churn is heavily front-loaded, with customers in their first three months showing a churn rate of about 57%, more than double the rate seen later in the lifecycle, which reveals a clear pattern where those who make it past the early stage tend to stay, making early-stage customers both the most vulnerable and the most critical group to retain.
Stage 2 — Root-Cause Segmentation: Why and Who Is Leaving
Problem Statement: After identifying when churn occurs, the next step is to understand why customers leave. The goal is to focus on identifying the key structural factors that consistently drive higher risk of attrition.
Analytical Approach: The analysis focuses on three dimensions, contract type, internet service category, and payment method, examined in parallel using an unpivoted dataset to keep comparisons consistent. Within each tenure group, churn rate and churn share are calculated to show which segments drive the most attrition. Window functions are then applied to capture how churn is distributed within each group, highlighting concentration rather than simple counts.
Key Assumption:
| Assumption | Rationale |
|---|---|
| Contract structure influences churn | Customers with short-term agreements face fewer constraints, making it easier for them to leave. |
| Service expectations affect retention | Customers paying for premium services expect better performance, and when that expectation isn’t met, churn increases. |
| Payment experience shapes early behavior | Complicated or manual payment methods introduce friction, which increases the likelihood of early drop-off. |
Key Findings: The analysis highlights three main drivers behind churn.
First, contract type stands out as the strongest factor, with month-to-month customers driving most of the churn across all stages. With no long-term commitment, leaving becomes an easy decision.
Second, fiber optic customers show consistently higher churn, even though they pay more. This points to a gap between what customers expect and what they experience, suggesting the issue lies in perceived service quality, not pricing.
Third, payment methods play a role early on, especially those that require more effort, like bank transfers or mailed checks. This friction increases drop-off during the early stages, though its impact fades as customers settle into привычки.
Overall, churn is shaped by three forces: low commitment, gap expectations, and early friction in the customer experience.
Stage 3 — Value-at-Risk Scoring: Who Is Worth Saving
Problem Statement: Not all churn carries the same financial weight. This step focuses on identifying which customers should be prioritized based on their potential impact on revenue?, introducing a financial lens into the segmentation process.
Analytical Approach: A value-at-risk (VaR) score is calculated for each customer by combining their lifetime value with their churn probability, capturing both how much revenue is at stake and how likely it is to be lost. Customers are then grouped across two dimensions, value (High: ≥$3,500 vs Low: <$3,500) and risk (High Risk: churn score ≥60 vs Low Risk: <60), forming a four-segment prioritization matrix. VaR is calculated both at the total segment level and per customer, allowing for clear comparisons in terms of overall impact and individual risk.
A value-at-risk (VaR) score is calculated: VaR = Customer Lifetime Value × Churn Probability
Key Assumption:
| Assumption | Detail |
|---|---|
| CLTV threshold | A cutoff of $3,500 separates high- and low-value customers. This is a business-defined benchmark and should be reviewed over time as customer value changes. |
| Churn score threshold | A score of 60 marks the boundary for high-risk customers, representing a clear step above the average churn likelihood. |
| VaR linearity | The model assumes a direct relationship between churn probability and expected revenue loss, which keeps the calculation simple but does not account for potential interaction effects. |
Key Findings: The VaR analysis shows that customers with 36++ months of tenure carry the highest total value at risk, driven by their accumulated lifetime value combined with meaningful churn probabilities. On a per-customer level, early-stage users stand out, as their risk is high relative to their value, making them more responsive to targeted intervention.
The most important segment is the High Value / High Risk group (CLTV ≥ $3,500 and churn score ≥ 60), which represents the top retention priority due to both a high likelihood of churn and significant financial impact. In contrast, the High Value / Low Risk group is less urgent but still important, offering an opportunity to strengthen loyalty and prevent future risk among already valuable customers.
Stage 4 — Action Plan & ROI Layer: Translating Insight into Action
Problem Statement: Insights on their own don’t create value unless they lead to action. This final step turns analysis into a clear execution plan by defining which customers to target, what actions to take, and what return to expect. The focus shifts from analysis to prioritization and practical decision-making, using an ROI framework to treat retention spend as an investment, held to the same standards as any other business
Analytical Approach: Customers who have not churned are grouped into four priority tiers based on their value, risk level, contract type, and revenue contribution. A minimum monthly charge of $30 is applied to filter out low-value accounts, ensuring focus stays on segments with meaningful impact. For each tier, ROI is calculated by estimating the value that can be saved from intervention and comparing it against the cost, using expected value saved (Value at Risk × Save Rate) and the corresponding return on investment formula.
ROI is calculated using:
- Expected Value Saved = Value at Risk × Save Rate
- ROI = (Value Saved − Cost) / Cost
Priority Tier:
| Priority Tier | Strategic Action | Target Customer Segment | Average Cost per Customer | Average Save Rate | ROI |
|---|---|---|---|---|---|
| P1 | Save Now | High Value + High Risk + M2M (Unit Economy > 30) | $15 | 30% | 67.1× |
| P2 | Prevent Churn | High Value + Low Risk + M2M (Unit Economy > 30) | $10 | 20% | 38.5× |
| P3 | Protect & Upsell | High Value + Contracted 1Y/2Y (Unit Economy > 30) | $6 | 12% | 49.9× |
| P4 | Monitor | Others | — | — | — |
ROI SUMMARY:
| Customer Priority Tier | Customers | Value at Risk ($) | Expected Saved ($) | Total Cost ($) | ROI (×) |
|---|---|---|---|---|---|
| P1 — Save Now | 341 | 1,161,810 | 348,543 | 5,115 | 67.1× |
| P2 — Prevent Churn | 700 | 1,384,186 | 276,837 | 7,000 | 38.5× |
| P3 — Protect & Upsell | 1,904 | 4,841,650 | 580,998 | 11,424 | 49.9× |
| TOTAL (P1–P3) | 2,945 | 7,387,645 | 1,206,378 | 23,539 | ~51× |
Key Findings:
- The ROI across all three priority tiers is strong and clearly justifies action. The P1 segment, made up of 341 high-value, high-risk, month-to-month customers, delivers a projected ROI of 67.1× with a relatively low total cost of $5,115. This reflects how much value is concentrated in a small but critical group of customers who are at immediate risk of leaving.
- The P3 segment shows a different kind of opportunity. While each customer is less urgent, the group holds the largest total value at risk at $4.84M and still delivers a 49.9× ROI through low-cost, scalable interventions.
- Taken together, the P1 to P3 strategy targets 2,945 customers, protects $7.4M in value at risk, and is expected to recover $1.2M in value with a total cost of $23,539, resulting in an overall ROI of about 51×. This isn’t a small efficiency improvement, it’s a clear opportunity to drive meaningful impact using focused, cost-effective retention actions.
Results and Recommendations
To put this strategy into action, the focus should stay on three clear priorities, each tied to value, risk, and expected return.
- P1 – Save Now: With a 30% save rate at $15 per customer, this segment delivers 67.1× ROI. These customers are high value and at immediate risk, so speed matters. Use targeted discounts (10–20%), bill credits, and strong incentives to move them from month-to-month to a 1-year contract. This is the one segment where higher spend is justified, since the cost of losing them is far greater.
- P2 – Prevent Churn: At a 20% save rate and $10 cost, this group generates 38.5× ROI. The goal here is to act before risk increases. Offer light incentives for contract upgrades and bundle deals, such as internet with add-ons, to increase commitment without overspending.
- P3 – Protect & Upsell: With a 12% save rate at $6 cost, this segment delivers 49.9× ROI. These customers are already stable, so the focus shifts from retention to growth. Upsell premium plans, introduce add-ons, and offer loyalty perks to strengthen engagement and increase lifetime value.
- P4 – Monitor: This group doesn’t justify active investment. With no clear return, the right approach is to monitor behavior without allocating budget, keeping resources focused on higher-impact segments.
For a deeper dive into the methodologies and data, the project’s GitHub repository is available at GitHub Repository.