1.3: Running Case Study - Mobile App Customer Churn Analysis

Running Case Study: Mobile App Customer Churn Analysis

Throughout this section, we'll follow MobileFlow, a subscription-based fitness app company, as they investigate: "What factors are driving customer churn in Q4? [6]"

---

Stage 1: Define Goals and Requirements

Before any data is gathered and prepared, our first priority and objective will be to define the problem or question that needs to be answered. In this initial phase, an analyst or team of analysts can clarify the business objectives and determine what data and analysis are required to address them. A well-defined question will guide the entire analytics process and ensure the results are relevant and actionable.

MobileFlow Case Study:

The analytics team at MobileFlow starts with a clear question: "What factors are driving the 15% increase in customer churn we observed in Q4?" Success is defined as identifying at least three actionable factors that can reduce churn by 5% in the next quarter.

---

Stage 2: Data Collection (and Access)

In this phase, the second priority revolves around the relevant data gathered from various sources. Data can come from internal systems (databases, transaction records, CRM systems), user-generated content, sensors/IoT devices, surveys, or external open data sources – whichever are needed to answer the question. The data may be structured (organized in tables) or unstructured (text, images, etc.). Often, this step involves pulling data from multiple databases or APIs and may require setting up pipelines to continually collect incoming data.

Key Insight: Data collection is an ongoing and dynamic process; new data is constantly generated, so organizations frequently need to automate and continuously manage data inflows.

MobileFlow Case Study:

The team collects user subscription data from their billing system, in-app activity logs from their mobile analytics platform, customer support ticket information, and user demographic data from their registration database.

---

Stage 3: Data Storage and Management

Once collected, the data must be stored in an organized and secure manner. In many cases, raw data is extracted, transformed, and loaded into a centralized repository such as a data warehouse or data lake (this process is known as ETL/ELT – Extract, Transform, Load). Proper storage ensures that data is in a usable format and that different data sources can be integrated. This stage also involves setting up databases or cloud storage solutions and managing data access. Good data management lays the foundation for efficient analysis by providing a single source of truth and maintaining data quality.

Scale Considerations: In smaller-scale analyses, this storage step might be as simple as gathering data into a spreadsheet or local database, whereas in big data scenarios it involves significant engineering infrastructure.

MobileFlow Case Study:

MobileFlow loads all collected data into their cloud-based data warehouse using an automated ETL pipeline that runs nightly, ensuring all departments access the same unified dataset.

---

Stage 4: Data Cleaning and Preprocessing

Raw data is often messy – it can contain errors, missing values, duplicates, or inconsistencies. In this step, data is cleaned and prepared to improve its quality. This includes handling missing data, correcting errors or outliers, standardizing formats, and combining datasets if multiple sources exist. Data cleaning (also called data scrubbing or preprocessing) is critical because properly cleaned data ensures accurate and reliable results. Analysts may remove corrupt records, reconcile differences (for example, ensuring consistent units or categories), and filter out irrelevant information.

Critical Point: This step can be time-consuming, but it is essential to avoid garbage-in, garbage-out issues in analysis.

MobileFlow Case Study:

The team discovers missing activity logs for 8% of users and inconsistent date formats across different systems. They handle missing data through interpolation methods and standardize all timestamps to UTC format.

---

Stage 5: Data Analysis and Modeling

After preparation, the analysis phase begins. Here, analysts explore the data to uncover insights. This may start with exploratory data analysis (EDA) – using statistical summaries and visualization to understand distributions, correlations, and anomalies in the data. EDA helps in hypothesis generation and selection of modeling approaches. Following exploration, analysts may apply more formal analytical modeling techniques. This could be as simple as calculating metrics and aggregating data to answer descriptive questions, or as complex as training machine learning models for prediction. Techniques such as regression analysis, classification, clustering, or time-series forecasting are employed in this phase to extract deeper insights from the data.

MobileFlow Case Study:

EDA reveals that users who engage with the app's social features less than twice per week show significantly higher churn rates. The team builds a logistic regression model to quantify this relationship and identify other predictive factors.

---

Stage 6: Visualization and Communication

Once analysis yields results, the insights must be interpreted and communicated. This stage involves summarizing the key findings and presenting them in a clear, accessible way, often using data visualization. Creating charts, graphs, and dashboards helps stakeholders understand the patterns or predictions derived from the data. Effective communication is vital: the goal is to translate analytical results into a compelling narrative or a set of recommendations for decision-makers. Techniques from data visualization and storytelling are used to highlight the most relevant insights.

Best Practice: The results of data analytics should be formatted so that they're accessible to and understandable by others, particularly decision-makers in a company.

MobileFlow Case Study:

The analytics team creates an interactive dashboard showing that users who engage with social features at least twice weekly are 40% less likely to churn. They present this finding alongside recommendations to the product and marketing teams.

---

Stage 7: Insight to Action

In this final phase, the organization acts on the insights obtained. This could mean making a strategic decision, implementing a change in business operations, launching a targeted marketing campaign, etc., based on the data-driven recommendations. Often, the analytics lifecycle is iterative – the outcome of one analysis may lead to new questions or require monitoring the effects of decisions, which brings the process back to the beginning. Modern analytics workflows acknowledge this by incorporating feedback loops and ongoing monitoring of key metrics.

The Iterative Nature: The ability to swiftly go from data to insight to action (and then measure the results) is a hallmark of data-driven organizations.

MobileFlow Case Study:

MobileFlow launches a targeted engagement campaign to encourage social feature usage among at-risk users. MobileFlow also modifies their onboarding process to highlight social features. The team sets up monitoring to track the campaign's impact on churn rates, planning to revisit their analysis in 30 days to measure effectiveness.

---

Key Takeaways

It's important to note that while the stages above are presented sequentially for clarity, real-world data analytics is rarely strictly linear. Teams may revisit earlier steps as new data becomes available or as initial findings prompt new hypotheses. For example, during analysis one might discover additional data is needed, requiring a return to the data collection phase. Despite these iterations, the overarching lifecycle provides a structured framework for transforming raw data into meaningful insight and ultimately into informed action.

Figure \(\PageIndex{1}\): A timeline traces how the MobileFlow fitness app team worked through each stage of the data lifecycle — from asking why users were leaving to launching a campaign that kept them engaged. (CC BY 4.0; Felix Amoruwa) Access to a detailed description of Figure \(\PageIndex{1}\)