1.6.4: Case Study Analysis
- Page ID
- 58284
\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)
\( \newcommand{\dsum}{\displaystyle\sum\limits} \)
\( \newcommand{\dint}{\displaystyle\int\limits} \)
\( \newcommand{\dlim}{\displaystyle\lim\limits} \)
\( \newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\)
( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\)
\( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)
\( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\)
\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)
\( \newcommand{\Span}{\mathrm{span}}\)
\( \newcommand{\id}{\mathrm{id}}\)
\( \newcommand{\Span}{\mathrm{span}}\)
\( \newcommand{\kernel}{\mathrm{null}\,}\)
\( \newcommand{\range}{\mathrm{range}\,}\)
\( \newcommand{\RealPart}{\mathrm{Re}}\)
\( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)
\( \newcommand{\Argument}{\mathrm{Arg}}\)
\( \newcommand{\norm}[1]{\| #1 \|}\)
\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)
\( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\AA}{\unicode[.8,0]{x212B}}\)
\( \newcommand{\vectorA}[1]{\vec{#1}} % arrow\)
\( \newcommand{\vectorAt}[1]{\vec{\text{#1}}} % arrow\)
\( \newcommand{\vectorB}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vectorC}[1]{\textbf{#1}} \)
\( \newcommand{\vectorD}[1]{\overrightarrow{#1}} \)
\( \newcommand{\vectorDt}[1]{\overrightarrow{\text{#1}}} \)
\( \newcommand{\vectE}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash{\mathbf {#1}}}} \)
\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\(\newcommand{\longvect}{\overrightarrow}\)
\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)
\(\newcommand{\ket}[1]{\left| #1 \right>}\)
\(\newcommand{\bra}[1]{\left< #1 \right|}\)
\(\newcommand{\braket}[2]{\left< #1 \vphantom{#2} \right| \left. #2 \vphantom{#1} \right>}\)
\(\newcommand{\braopket}[3]{\left< #1 \vphantom{#2}\vphantom{#3} \right| #2 \vphantom{#1}\vphantom{#3} \left| #3 \vphantom{#1}\vphantom{#2} \right>}\)
\(\newcommand{\qmvec}[1]{\mathbf{\vec{#1}}}\)
\(\newcommand{\op}[1]{\hat{\mathbf{#1}}}\)
\(\newcommand{\expect}[1]{\langle #1 \rangle}\)
\(\newcommand{\dfn}[1]{\emph{\textbf{#1}}}\)
MobileFlow: From Data to Action on Customer Churn
MobileFlow is a subscription-based fitness app with 500,000 active users. In Q4, the analytics team noticed a concerning trend: customer churn had increased from 8% to 15% — nearly doubling. The CEO asked the data team to investigate and recommend actions.
The team followed the data-to-insight lifecycle:
Stage 1 — Define Goals: The team framed the question: "What factors are driving the 15% churn rate in Q4, and how can we reduce it to below 10%?"
Stage 2 — Collect Data: They gathered data from four sources: billing system (subscription status, payment history), in-app analytics platform (feature usage, session frequency), customer support tickets (complaints, cancellation reasons), and user demographics from the registration database.
Stage 3 — Store: All data was loaded into the company's cloud data warehouse using an automated nightly ETL pipeline.
Stage 4 — Clean: The team found 8% of users had missing activity logs and inconsistent date formats across systems. They used interpolation for missing data and standardized timestamps to UTC.
Stage 5 — Analyze: Exploratory analysis revealed that users engaging with social features fewer than twice per week had significantly higher churn. A logistic regression model confirmed this and identified two additional factors: users who hadn't updated the app in 60+ days and users whose workout frequency dropped by more than 50%.
Stage 6 — Visualize: The team built an interactive dashboard showing churn risk scores by user segment, the relationship between social feature usage and retention, and a geographic breakdown of churn patterns.
Stage 7 — Act: Based on findings, MobileFlow launched a targeted push notification campaign encouraging social feature usage among at-risk users, redesigned the onboarding flow to highlight social features, and implemented automatic app update reminders. They set up monitoring to track impact over 30 days.
Results: After 60 days, churn decreased from 15% to 9.5%, exceeding the target. Social feature engagement increased by 35% among the targeted segment.
Discussion Questions:
-
Map each action MobileFlow took to one of the four analytics types (descriptive, diagnostic, predictive, prescriptive). Were all four types represented in their approach? Which type provided the most value, and why?
-
The team discovered that 8% of users had missing activity logs. How might the reason for this missing data affect the analysis? What if the missing logs were specifically from users who had already churned (MNAR — Missing Not at Random)?
-
The logistic regression model identified social feature usage, app update recency, and workout frequency decline as churn predictors. What other factors might be worth investigating? How would you design a follow-up analysis to validate whether the interventions actually caused the churn reduction (vs. other factors)?
-
MobileFlow's approach was reactive — they investigated after churn had already increased. How could MobileFlow redesign their analytics approach to be proactive, detecting churn risk before it manifests in actual cancellations? What real-time data and models would be needed?