Business Framework Interactive

Cohort Analysis

Group users by signup date and track behavior over time. Essential for understanding retention, LTV evolution, and comparing user groups.

📊 What is Cohort Analysis?

Group users by when they joined (cohort), then track a metric over time. This reveals patterns hidden in aggregate data.

Why It Matters

• Aggregate retention can improve while cohort retention worsens (Simpson's paradox)
• Different cohorts have different behaviors
• See true product improvements over time

Cohort

Jan

70%

55%

45%

Feb

65%

50%

Mar

75%

Retention Settings

40 90

30 80

20 70

2 15

📊 Cohort LTV

January Cohort $NaN

February Cohort $NaN

March Cohort $NaN

LTV = Σ (retention × monthly revenue)

Retention Curves by Cohort

Each line = one cohort's retention journey. Compare cohorts to identify improvements.

Month 3 Retention Comparison

🎯 Use Cases

Retention Curves

Track how each signup cohort retains over time

A/B Test Impact

Compare cohorts with different treatments

Seasonality

Understand if summer vs winter signups behave differently

LTV by Acquisition

Which channels bring highest LTV cohorts?

🏀 Sports Betting Applications

Acquisition Quality

→ Compare LTV by acquisition channel
→ Identify which campaigns bring quality users
→ Track promo impact on future behavior

Product Improvements

→ Did new feature improve retention?
→ Compare pre/post cohorts fairly
→ Seasonality analysis (NFL season vs off-season)

R Code Equivalent

# Cohort analysis
library(dplyr)
library(tidyr)

create_cohort_table <- function(users, activity) { 
  users %>%
    mutate(cohort = floor_date(signup_date, "month")) %>%
    left_join(activity, by = "user_id") %>%
    mutate(months_since = interval(cohort, activity_date) %/% months(1)) %>%
    group_by(cohort, months_since) %>%
    summarise(
      active = n_distinct(user_id),
      .groups = "drop"
    ) %>%
    left_join(
      users %>%
        mutate(cohort = floor_date(signup_date, "month")) %>%
        group_by(cohort) %>%
        summarise(cohort_size = n()),
      by = "cohort"
    ) %>%
    mutate(retention = active / cohort_size * 100)
}

# Calculate cohort LTV
cohort_ltv <- function(retention_curve, avg_monthly_rev) { 
  sum(retention_curve * avg_monthly_rev / 100)
}

# Example
retention <- c(100, 70, 55, 45)
ltv <- cohort_ltv(retention, 25)
cat(sprintf("Cohort LTV: $%.0f\n", ltv))

✅ Key Takeaways

• Group users by signup period (cohort)
• Track metrics over time for each cohort
• Reveals patterns hidden in aggregate data

• Compare cohorts to measure improvement
• Essential for LTV and retention analysis
• Identify acquisition channel quality