Finding the High-Risk Moments in Your Customer Journey
- Gandhinath Swaminathan
- Sep 22
- 6 min read
In my previous articles, I've discussed lagging indicators of customer churn. These are metrics that look backward.
Net Revenue Retention (NRR) shows you the total change in recurring revenue from a group of customers over time, factoring in upgrades, downgrades, and churn.
Churn Rate is simpler. It's the percentage of customers who left in a given period.
These numbers are useful, but they only tell you what already happened.
A survival curve doesn't just show that customers are leaving; it shows when. The steep drops in this chart represent high-risk periods—the precise moments where proactive retention efforts can make the most difference.
Growth isn't just about acquisition. It's about understanding the natural lifecycle of your customer relationships. Seeing that lifecycle clearly is the first step toward improving it.
A Better Tool for Understanding Time
Survival Analysis is a statistical method that models time-to-event data. It originated in medical research to estimate patient lifespans but is perfectly suited for business. In our context, it models a customer's 'lifespan' with your company.
It answers the question:
"What is the probability that a customer will stay with us for a specific amount of time?"
This is different from a standard churn rate. A monthly churn rate gives you a single, static number. A survival analysis gives you a curve, showing how the probability of a customer staying with your company (subscription) changes over time. This curve exposes the high-risk periods in the customer journey.
Common Methods for Survival Analysis
There are three main families of survival analysis models. Each offers a different level of detail.
Parametric Models assume that the customer survival time follows a known statistical distribution (like a bell curve). These models are powerful when you have a good reason to believe your customer behavior fits a specific pattern.
Non-Parametric Models make no assumptions about the underlying distribution of survival times. They are flexible and provide a direct look at the data you have. The Kaplan-Meier model is in this family.
Semi-Parametric Models are a hybrid. The most common is the Cox Proportional-Hazards model. It does not assume a specific distribution for survival time but allows you to analyze how different variables—like pricing plans or usage frequency—affect the churn risk.
For this discussion, we will focus on the simplest and most direct method: the non-parametric Kaplan-Meier estimator. It requires fewer statistical assumptions. It offers a clear, direct view of customer survival over time. For most businesses, it is a good starting point since it provides immediate, actionable information.
The Story Told by the Kaplan-Meier Curve
The Kaplan-Meier curve is a graph that plots the probability of customer retention over time. The vertical axis shows the percentage of customers who are still active. The horizontal axis tracks time in days, months, or years.
The curve always starts at 100%. As time moves forward and customers churn, the curve steps down.
A steep drop signals a high-risk period where many customers are leaving.
A flat section indicates a period of strong retention.
What if you could stop asking better questions and finally see the answers hiding in your data? Imagine pinpointing the exact moment, the costly blind spots where customer loyalty erodes and revenue is lost, as in:
The critical drop-off after onboarding friction.
The invisible wall they hit at the six-month mark when your customer is not realizing the perceived value.
The sudden spike in risk just before renewal.
The story of when is revealed by the shape of a single, powerful curve that charts the path from intervention to outcome
A Practical Guide: Building the Curve in a Spreadsheet
The goal is to create a chart that clearly shows you when your customers are most likely to churn.
You can build your first survival model using common spreadsheet applications like Microsoft Excel or Google Sheets. Note that there is no need for specialized software to begin.
We will track a cohort of 10 customers over one year. For each customer, you need two data points:
Tenure: The number of days the customer was with your company.
Churn Status: A flag to show if the churn event was observed. We use 1 if the customer churned and 0 if they were still active at the end of the observation period. In statistics, this is called "censoring." It is important because it allows us to include data from customers who have not yet churned, giving us a more complete picture of survival.
Step 1: The Data
Create a simple table in your spreadsheet. Let's assume your data is in a table named ChurnData.
Customer ID | Tenure (Days) | Churned (Event Observed) |
1 | 35 | 1 |
2 | 68 | 1 |
3 | 92 | 1 |
4 | 120 | 0 |
5 | 154 | 1 |
6 | 188 | 0 |
7 | 211 | 1 |
8 | 245 | 0 |
9 | 280 | 1 |
10 | 365 | 0 |
Step 2: The Calculation Table
Set up a new table to calculate the survival probability. This table tracks the number of customers at risk and their likelihood of survival at each churn event.
The table requires five columns:
Time (t): The specific days when a churn event occurred. List the unique Tenure days where Churned is 1.
35, 68, 92, 154, 211, 280
Number at Risk (n): The number of active customers just before time (t). For the first time point (day 35), all 10 customers are at risk. For each row after, subtract the customers who churned or were censored (left the observation group while still active) in the prior period.
At t=35, n=10.
At t=68, n=9 (customer 1 churned).
At t=92, n=8 (customer 2 churned).
At t=154, n=6 (customer 3 churned, customer 4 was censored at day 120).
At t=211, n=4 (customer 5 churned, customer 6 was censored at day 188).
At t=280, n=2 (customer 7 churned, customer 8 was censored at day 245).
Number of Events (d): The number of customers who churned at time (t). In this example, it is always 1.
Survival Probability (S(t)): The probability of a customer surviving that specific time point. This is the survival rate for each specific period. The formula is (n-d)/n.
For t=35: (10-1)/10 = 0.90
For t=68: (9-1)/9 = 0.89
For t=92: (8-1)/8 = 0.88
For t=154: (6-1)/6 = 0.83
For t=211: (4-1)/4 = 0.75
For t=280: (2-1)/2 = 0.50
Cumulative Survival: This is the Kaplan-Meier estimate. It is the probability of a customer having survived from the start until time (t). Calculate this by multiplying the current period’s survival probability by the previous period’s cumulative survival.
For t=35: 0.90
For t=68: 0.90 * 0.89 = 0.80
For t=92: 0.80 * 0.88 = 0.70
For t=154: 0.70 * 0.83 = 0.58
For t=211: 0.58 * 0.75 = 0.44
For t=280: 0.44 * 0.50 = 0.22
Step 3: The Formulas (Optional - included here for completeness)
Time (t):
=SORT(UNIQUE(FILTER(ChurnData[Tenure (Days)], ChurnData[Churned (Event Observed)]=1)))Number at Risk (n):
For the first time point (e.g., in cell E2), the formula would be =COUNTIF(ChurnData[Tenure (Days)], ">="&E2).Number of Events (d):
For the time point in cell E2, the formula is =COUNTIFS(ChurnData[Tenure (Days)], E2, ChurnData[Churned (Event Observed)], 1).Survival Probability (S(t)):
The formula is (n-d)/n. If your 'n' is in column F and 'd' is in column G, the formula is =(F2-G2)/F2.Cumulative Survival:
If cumulative survival is in column I, the first cell is =H2 and the next cell down is =I2*H3.Step 4: Final Table & The Chart
Your final table in the spreadsheet will look like this:
Time (t) | Number at Risk (n) | Number of Events (d) | Survival Probability (S(t)) | Cumulative Survival |
35 | 10 | 1 | 0.90 | 0.90 |
68 | 9 | 1 | 0.89 | 0.80 |
92 | 8 | 1 | 0.88 | 0.70 |
154 | 6 | 1 | 0.83 | 0.58 |
211 | 4 | 1 | 0.75 | 0.44 |
280 | 2 | 1 | 0.50 | 0.22 |
You can create a "Scatter with Straight Lines" chart.
X-Axis: Time (t)
Y-Axis: Cumulative Survival
The resulting graph is the Kaplan-Meier Curve. It is a step-down function that visualizes the decay of your customer base over time, highlighting the specific points where intervention may be most effective. You can now see that after 280 days, only 22% of the initial group is expected to remain. The drops show you exactly where to look for problems in your customer experience.
From Insight to Action
This analysis gives you a map of your customer lifecycle. It moves you from looking at a single, historical churn number to understanding the forward-looking risk at every stage. You now know when to act (intervention). The immediate strategic imperative is to quantify the impact on your growth model.
The critical next step is to translate that churn forecast into a precise customer acquisition target. In the next post, we'll break down the exact formula every growth leader needs—the number of new customers required to outpace churn and hit your revenue targets. This precision will shift your strategy from hopeful to confident execution. Stay tuned.
Comments