
Product analytics case study
Fumble Subscription Cohort Analysis
Comparing six-month retention across Standard and Pro subscription cohorts.
I analyzed July to December 2018 monthly cohorts to compare retention behavior across Fumble’s two subscription plans and surface plan-level patterns relevant to an investor discussion.
- monthly cohorts
- 6
- subscription plans
- 2
- analysis window
- 6-month
- users in Jul–Dec cohorts
- 2,013
What the brief asked
Fumble is a dating app with two subscription plans, Standard and Pro. The exercise asked for the July to December 2018 monthly cohorts to be analyzed over their first six months for retention and average revenue, with the results prepared for an investor conversation.
This is a product exercise on a supplied dataset. It reports what the data shows and does not claim why, and no product change was shipped.
From subscriber rows to cohort tables
Six monthly cohorts per plan, tracked for months 0 to 5. A subscriber counts as retained at month m when their active months exceed m; month 0 is the cohort size.
- Raw subscription data
4,711 subscribers with plan, start month and active months.
- Cohort assignment
Subscribers grouped by plan and start month, Jul to Dec 2018.
- Active-month calculation
Retained at month m when active months exceed m.
- Cohort retention table
Counts and percentages for months 0 to 5, built in Python.
- Plan comparison
Standard and Pro compared cohort by cohort at month 5.
The charts on this page are computed from the current workbook. The original one-page analysis was built from an earlier data snapshot and differs slightly for the July cohort (Standard 145 vs 151 users, Pro 84 vs 85), so its numbers are kept out of these charts rather than mixed in.
The brief also asked for average revenue. The workbook holds a lifetime revenue total per subscriber with no stated first-six-month method, so no revenue figure is shown rather than assuming one.
Standard Plan: 1,391 users across six cohorts
| Retention month | |||||||
|---|---|---|---|---|---|---|---|
| Cohort | Users | 0 | 1 | 2 | 3 | 4 | 5 |
| Jul ’18 | 151 | 100.0 | 90.1 | 80.8 | 68.2 | 60.9 | 57.6 |
| Aug ’18 | 160 | 100.0 | 89.4 | 73.8 | 63.8 | 57.5 | 54.4 |
| Sep ’18 | 234 | 100.0 | 82.9 | 80.3 | 65.8 | 60.3 | 46.6 |
| Oct ’18 | 279 | 100.0 | 90.7 | 75.3 | 64.9 | 41.6 | 25.1 |
| Nov ’18 | 260 | 100.0 | 75.8 | 72.3 | 51.9 | 22.3 | 21.5 |
| Dec ’18 | 307 | 100.0 | 95.4 | 64.5 | 16.0 | 15.3 | 13.4 |
Values are % of the cohort. Darker tint means a higher value.
- Standard Plan, all six cohorts pooled
- Individual cohorts
Pooled line sums all six cohorts; faint lines are the individual cohorts.
| Retention month | Jul | Aug | Sep | Oct | Nov | Dec | Standard Plan, all six cohorts pooled |
|---|---|---|---|---|---|---|---|
| 0 | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% |
| 1 | 90.1% | 89.4% | 82.9% | 90.7% | 75.8% | 95.4% | 87.4% |
| 2 | 80.8% | 73.8% | 80.3% | 75.3% | 72.3% | 64.5% | 73.6% |
| 3 | 68.2% | 63.8% | 65.8% | 64.9% | 51.9% | 16.0% | 52.0% |
| 4 | 60.9% | 57.5% | 60.3% | 41.6% | 22.3% | 15.3% | 39.3% |
| 5 | 57.6% | 54.4% | 46.6% | 25.1% | 21.5% | 13.4% | 32.4% |
Pro Plan: 622 users across six cohorts
| Retention month | |||||||
|---|---|---|---|---|---|---|---|
| Cohort | Users | 0 | 1 | 2 | 3 | 4 | 5 |
| Jul ’18 | 85 | 100.0 | 96.5 | 92.9 | 88.2 | 72.9 | 71.8 |
| Aug ’18 | 85 | 100.0 | 98.8 | 92.9 | 85.9 | 76.5 | 72.9 |
| Sep ’18 | 95 | 100.0 | 92.6 | 86.3 | 76.8 | 65.3 | 53.7 |
| Oct ’18 | 101 | 100.0 | 98.0 | 85.1 | 71.3 | 60.4 | 44.6 |
| Nov ’18 | 131 | 100.0 | 78.6 | 72.5 | 54.2 | 29.0 | 29.0 |
| Dec ’18 | 125 | 100.0 | 99.2 | 72.0 | 29.6 | 28.0 | 25.6 |
Values are % of the cohort. Darker tint means a higher value.
- Pro Plan, all six cohorts pooled
- Individual cohorts
Pooled line sums all six cohorts; faint lines are the individual cohorts.
| Retention month | Jul | Aug | Sep | Oct | Nov | Dec | Pro Plan, all six cohorts pooled |
|---|---|---|---|---|---|---|---|
| 0 | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% |
| 1 | 96.5% | 98.8% | 92.6% | 98.0% | 78.6% | 99.2% | 93.2% |
| 2 | 92.9% | 92.9% | 86.3% | 85.1% | 72.5% | 72.0% | 82.2% |
| 3 | 88.2% | 85.9% | 76.8% | 71.3% | 54.2% | 29.6% | 64.5% |
| 4 | 72.9% | 76.5% | 65.3% | 60.4% | 29.0% | 28.0% | 51.9% |
| 5 | 71.8% | 72.9% | 53.7% | 44.6% | 29.0% | 25.6% | 46.5% |
Pro cohorts retain more at month 5
- Pro Plan
- Standard Plan
Share of each cohort still active at retention month 5, from the current workbook.
| Category | Pro Plan | Standard Plan |
|---|---|---|
| Jul ’18 | 71.8% | 57.6% |
| Aug ’18 | 72.9% | 54.4% |
| Sep ’18 | 53.7% | 46.6% |
| Oct ’18 | 44.6% | 25.1% |
| Nov ’18 | 29.0% | 21.5% |
| Dec ’18 | 25.6% | 13.4% |
- Pro-plan cohorts have higher month-5 retention than Standard-plan cohorts in each of the six Jul to Dec cohorts.For example Jul 71.8% vs 57.6%, Dec 25.6% vs 13.4%.
- Standard-plan cohorts contain more users in absolute terms.1,391 Standard vs 622 Pro users across the six cohorts.
These are descriptive observations. The data does not explain why Pro users stay longer, and later cohorts have had less time to accumulate months, so a cohort-by-cohort read is the fairest comparison.
Related design prototype
Tumble: Report a user
A separate design exercise from the same course file, shown as related product-design work. It is not part of the cohort analysis.
1 / 4
Step 01
Home: report entry
A profile card in the Tumble dating app carries a Report button next to the person's name.
- Entry pointReport beside the profile name
- Also on screenSuper Like and Mode
- Bottom barProfile, Bees, Likes, Chat
Step 02
Report Profile
Tapping Report opens a form with two dropdowns and a place to record a statement.
- Question 1What is wrong with this profile?
- Question 2What kind of action do you want to take?
- StatementRecord your statement
Step 03
Filled report
With a reason, an action and a recorded statement, the form shows the statement and is ready to submit.
- Report reasonAbusive Content
- Desired actionBlock, Don't show profile
- Recorded statementPlay Statement, 2:00
Step 04
Grievance submitted
A confirmation sets an expectation for a reply and returns the user to the home screen.
- Expected response time48 hours
- Return pathBack to Home
Source dataset
PMF: Week 5 Project Data Set
The raw subscriber data behind every chart on this page.
| ID | Join Date | End Date | Plan Type | Revenue | Start Month | Active Months |
|---|---|---|---|---|---|---|
| 89950 | 2018-01-25 | 2019-03-02 | Standard Plan | $160.00 | Jan-2018 | 14 |
| 89951 | 2018-01-26 | 2018-03-27 | Standard Plan | $40.00 | Jan-2018 | 2 |
| 89952 | 2018-01-26 | 2021-01-27 | Pro Plan | $1,234.29 | Jan-2018 | 37 |
| 89953 | 2018-01-26 | 2021-01-27 | Pro Plan | $1,234.29 | Jan-2018 | 37 |
| 89954 | 2018-01-27 | 2021-01-27 | Pro Plan | $1,234.29 | Jan-2018 | 37 |
| 89955 | 2018-01-27 | 2021-07-11 | Standard Plan | $461.71 | Jan-2018 | 43 |
| 89958 | 2018-01-29 | 2018-10-26 | Standard Plan | $180.00 | Jan-2018 | 9 |
| 89984 | 2018-02-03 | 2018-04-19 | Standard Plan | $40.00 | Feb-2018 | 3 |
First 8 of 4,711 rows. The full workbook is not embedded in the page.