Many businesses spend enormous amounts of money acquiring customers.

However, a surprising truth in analytics is:

Acquiring a new customer is usually far more expensive than keeping an existing one.

A company that constantly acquires customers but loses them just as quickly is like pouring water into a leaking bucket.

This is why retention analytics is one of the most valuable areas in data science.

Learning Objectives

By the end of this lesson, you should be able to:

• Understand customer retention
• Understand customer churn
• Calculate retention rates
• Calculate customer lifetime value
• Estimate customer lifetime
• Understand cohort analysis
• Build simple churn prediction models
• Connect retention analytics to business decisions

Acquisition Versus Retention

Imagine two companies.

Company A

Acquires:

$$1000$$

customers every month.

Loses:

$$950$$

customers every month.

Net gain:

$$1000-950=50$$

customers.

Company B

Acquires:

$$300$$

customers every month.

Loses:

$$50$$

customers every month.

Net gain:

$$300-50=250$$

customers.

Although Company A acquires more customers, Company B grows faster.

Retention often matters more than acquisition.

What Is Retention?

Retention measures the percentage of customers who remain active.

Suppose a company starts with:

$$1000$$

customers.

After one year:

$$850$$

remain active.

Retention Rate is:

$$Retention\ Rate=\frac{850}{1000}=85\%$$

What Is Churn?

Churn measures the percentage of customers who leave.

Formula:

$$Churn\ Rate=\frac{Customers\ Lost}{Customers\ at\ Start}$$

Example:

Customers at start:

$$1000$$

Customers lost:

$$150$$

Then:

$$Churn=\frac{150}{1000}=15\%$$

Relationship Between Retention and Churn

Retention and churn are complements.

$$Retention=1-Churn$$

Example:

If:

$$Churn=15\%$$

then:

$$Retention=85\%$$

Why Churn Is Important

Suppose:

Customer Acquisition Cost:

$$CAC=200$$

Customer Lifetime Value:

$$LTV=1000$$

Looks good.

Now suppose churn suddenly doubles.

Customers leave sooner.

LTV falls.

The entire business model may become unprofitable.

Understanding Customer Lifetime

Suppose annual churn is:

$$10\%$$

A rough approximation for customer lifetime is:

$$Lifetime=\frac{1}{Churn}$$

Thus:

$$Lifetime=\frac{1}{0.10}=10\ years$$

Suppose churn rises to:

$$25\%$$

Then:

$$Lifetime=\frac{1}{0.25}=4\ years$$

The customer relationship becomes much shorter.

Deriving Customer Lifetime Value

A simple approximation is:

$$LTV=Revenue\ Per\ Period\times Customer\ Lifetime$$

Suppose:

Monthly revenue:

$$50$$

Average lifetime:

$$24\ months$$

Then:

$$LTV=50\times24=1200$$

Expected customer value:

$$$1200$$

Profit-Based Lifetime Value

A more realistic formula uses profit rather than revenue.

$$LTV=Profit\ Per\ Period\times Lifetime$$

Suppose:

Monthly profit:

$$20$$

Lifetime:

$$24\ months$$

Then:

$$LTV=20\times24=480$$

The customer generates:

$$$480$$

of profit.

The LTV:CAC Ratio

One of the most important metrics in business.

Formula:

$$LTV:CAC=\frac{LTV}{CAC}$$

Suppose:

$$LTV=1200$$

$$CAC=300$$

Then:

$$\frac{1200}{300}=4$$

Meaning:

Every dollar spent acquiring customers returns four dollars in lifetime value.

Common Benchmarks

Poor:

$$LTV:CAC<1$$

Break-even:

$$LTV:CAC=1$$

Healthy:

$$LTV:CAC>3$$

Excellent:

$$LTV:CAC>5$$

Cohort Analysis

A cohort is a group of customers who entered during the same period.

Example:

Instead of analyzing everyone together, we analyze each cohort separately.

Example Cohort Table

Suppose January acquired:

$$1000$$

customers.

Retention percentages:

Month 1:

$$\frac{900}{1000}=90\%$$

Month 2:

$$\frac{850}{1000}=85\%$$

Month 3:

$$\frac{800}{1000}=80\%$$

Month 4:

$$\frac{760}{1000}=76\%$$

Why Cohort Analysis Matters

Without cohorts, a company may think performance is improving simply because new customers are arriving.

Cohorts reveal:

• Customer quality
• Retention trends
• Marketing effectiveness
• Product improvements

Healthcare Example

Suppose a hospital launches a diabetes monitoring program.

Initial patients:

$$500$$

After one year:

$$425$$

remain enrolled.

Retention:

$$\frac{425}{500}=85\%$$

Churn:

$$\frac{75}{500}=15\%$$

The hospital can compare retention between clinics.

Supply Chain Example

Suppose a supplier has:

$$100$$

retail customers.

At year end:

$$92$$

continue ordering.

Retention:

$$\frac{92}{100}=92\%$$

Customer retention directly affects future revenue forecasts.

Churn Prediction

One of the most common data science problems is predicting who will leave.

Target variable:

$$Y=\begin{cases}
1,&\text{Customer Churns}\\
0,&\text{Customer Stays}
\end{cases}$$

Features might include:

• Number of purchases
• Average order value
• Days since last purchase
• Customer age
• Support interactions

Logistic Regression for Churn

A common model is logistic regression.

The probability of churn is:

$$P(Y=1|X)=\frac{1}{1+e^{-(\beta_0+\beta_1X_1+\cdots+\beta_pX_p)}}$$

Output:

A probability between:

$$0$$

and

$$1$$

Example

Suppose a model predicts:

$$P(Churn)=0.90$$

This customer is highly likely to leave.

Marketing may send:

• Discounts
• Loyalty rewards
• Personal outreach

to retain them.

Survival Analysis Perspective

Instead of asking:

“Will the customer leave?”

we ask:

“When will the customer leave?”

This leads to survival analysis.

Let:

$$T=Time\ Until\ Churn$$

We study:

$$P(T>t)$$

which is the probability the customer remains active after time $$t$$.

This becomes important in subscription businesses.

Python Example: Retention and Churn

customers_start = 1000
customers_end = 850
retention = customers_end / customers_start
churn = 1 - retention
print(f"Retention: {retention:.2%}")
print(f"Churn: {churn:.2%}")

Output:

Retention: 85.00%
Churn: 15.00%

Python Example: Customer Lifetime Value

monthly_revenue = 50
lifetime_months = 24
ltv = monthly_revenue * lifetime_months
print(f"LTV: ${ltv:.2f}")

Output:

LTV: $1200.00

Complete Business Example

Suppose:

• Advertising spend = $150,000
• New customers = 500
• Monthly revenue per customer = $60
• Annual churn = 20\%

Step 1: Calculate CAC

$$CAC=\frac{150000}{500}=300$$

Step 2: Estimate Lifetime

$$Lifetime=\frac{1}{0.20}=5\ years$$

Step 3: Annual Revenue

$$Revenue=60\times12=720$$

Step 4: Lifetime Value

$$LTV=720\times5=3600$$

Step 5: LTV:CAC Ratio

$$\frac{3600}{300}=12$$

This appears to be an extremely profitable acquisition strategy.

Key Takeaways

Retention is often more important than acquisition.

The most important formulas introduced in this lesson are:

$$Retention=\frac{Customers\ Remaining}{Customers\ at\ Start}$$

$$Churn=\frac{Customers\ Lost}{Customers\ at\ Start}$$

$$Retention=1-Churn$$

$$Lifetime=\frac{1}{Churn}$$

$$LTV=Revenue\ Per\ Period\times Lifetime$$

$$LTV:CAC=\frac{LTV}{CAC}$$

These concepts form the foundation of:

• Customer Analytics
• Marketing Analytics
• Subscription Analytics
• Healthcare Retention Programs
• Supply Chain Customer Management
• Churn Prediction Models

Exercises

1. A company starts with 2,000 customers and loses 300. Calculate retention and churn.
2. Monthly customer revenue is $80 and average lifetime is 30 months. Calculate LTV.
3. A company spends $400,000 and acquires 1,000 customers. Calculate CAC.
4. If annual churn is $$12\%$$, estimate customer lifetime.
5. Calculate the LTV:CAC ratio when:

$$LTV=2500$$

and

$$CAC=400$$

References

1. Gupta, Sunil & Lehmann, Donald. Managing Customers as Investments: The Strategic Value of Customers in the Long Run. Wharton School Publishing.
2. Fader, Peter. Customer Centricity: Focus on the Right Customers for Strategic Advantage. Wharton Digital Press.
3. Fader, Peter & Hardie, Bruce. Customer-Base Analysis in a Discrete-Time Noncontractual Setting. Marketing Science.
4. Gelman, Andrew et al. Bayesian Data Analysis. CRC Press.
5. James, Gareth, Witten, Daniela, Hastie, Trevor, Tibshirani, Robert. An Introduction to Statistical Learning. Springer.

Next Lesson

Lesson 3: Customer Segmentation — RFM Analysis, Behavioral Segmentation, and Clustering Methods for Data Scientists. This is where we begin identifying which customers are most valuable and how to target them using statistical and machine learning techniques.