Why Are We Losing Customers? A Telco Churn Exploration
A beginner-friendly exploratory data analysis (EDA) project built entirely in Excel, no coding required. This project walks through how to explore a real dataset, ask the right questions, and turn findings into a stakeholder-ready story.
If you're trying to break into data analytics and looking for a first project to build, feel free to fork/replicate this one,the full steps are below.
1.) The Business Question
Why are customers churning, and what should the business do about it?
This is the kind of open-ended question a stakeholder might actually ask,the goal of this project is to explore the data and turn it into a clear, actionable answer.
2.) Dataset
- Source: Telco Customer Churn from Kaggle - link : https://www.kaggle.com/code/farazrahman/telco-customer-churn-logisticregression
- File: WA_Fn-UseC_-Telco-Customer-Churn.csv
- Size: 7032 customer records (after removing rows with missing values)
- Key columns used: tenure, MonthlyCharges, Contract, Churn
- Tools Used Microsoft Excel (PivotTables, statistical functions, charts(no add-ins required))
3.) Data Cleaning Before analyzing, I checked for missing values and data type issues:
Found ~11 blank values in TotalCharges (new customers with 0 tenure),exclude them from calculations Verified Churn (Yes/No) and Contract fields were consistent, no whitespace or typos Checked for duplicate customerID rows, none found
4.) Key Findings
i) Contract type is the strongest churn signal Month-to-month customers churn at a dramatically higher rate than customers on 1-year or 2-year contracts. Flexibility for the customer means higher risk for the business?
ii) Churn happens early Half of all customers who churn do so within their first 10 months (median tenure), while customers who stay have a median tenure of 38 months (more than 3 years). There's also a solid group of long-term loyal customers who've stuck around for years (refer to the right end of histogram under non-churn), so the business has two very different types of customers, and they probably need different strategies to retain existing customers and to reduce churn rate.
| Count | Mean tenure | Median tenure | |
|---|---|---|---|
| Non-Churn | 5,163 | 37.65 months | 38.00 months |
| Churned | 1,869 | 17.98 months | 10.00 months |
iii) Churned customers pay more On average, customers who churn pay about $13/month more than those who stay ($74.44 vs $61.31), this indicates price sensitivity plays the role for customers already on flexible, no-commitment plans.
| Mean | Median | |
|---|---|---|
| Non-Churn | $61.31 | $64.45 |
| Churned | $74.44 | $79.65 |
5.) Recommendation
Retention efforts should prioritize two distinct groups, based on the data:
i) High-risk: new, month-to-month customers in their first year, especially those on higher-priced plans, this segment shows the highest combined churn risk. A targeted incentive (e.g. a discount for switching to an annual contract around month 3–6) could meaningfully reduce early churn.
ii) High-value: long-tenured customers (5+ years) this loyal segment represents low churn risk but high advocacy potential. Rather than discounting (which isn't needed to retain them), consider a loyalty recognition program e.g. an exclusive perk, early access to new plans/features, or a small "thank you" credit at each tenure milestone (3, 5, 7 years). Loyal customers are also more likely to refer others or leave positive reviews, so investing in this group can drive word-of-mouth growth, not just retention.
For those who interested to replicate this Project, please follow the steps below:
1.) Download the dataset from Kaggle (link above) and open it in Excel.
2.) Convert the range to a Table (Ctrl+T) for easier referencing.
3.) Check for missing values and inconsistent data types before analyzing anything.
4.) Build a PivotTable: Contract in Rows, Churn in Values (shown as % of Row Total) → this surfaces the #1 finding almost immediately.
5.) Get Mean and Median tenure by Churn status using a PivotTable (Average) plus a quick sort-and-MEDIAN() on the sorted blocks.
6.) Repeat step 5 for MonthlyCharges.
7.) Turn each finding into one simple chart, a bar chart, a histogram, or a grouped comparison chart.
8.) Write 2-3 sentences per finding explaining why it matters, not just what the number is.
9.) End with one clear, specific recommendation, this is what separates an analysis from a report.
📁 Files in This Repo eg-telco-churn-eda-public.xlsx, the working Excel file (PivotTables, formulas, and charts included) charts/ exported chart images used in this README
Part of my data analyst learning journey,follow bibyteswithme for more beginner-friendly projects :-)