An Excel-based analysis of customer churn and retention for a telecom provider, built on a publicly available Kaggle dataset. The project walks through the full analysis process — from raw data to pivot-table analysis to a single interactive dashboard — to help identify where and why customers are leaving.
Customer churn is a key retention challenge for telecom providers, and it’s rarely explained by a single factor. This project pulls a raw customer dataset into Excel, structures it for analysis using pivot tables, and builds out KPIs and visuals that surface where churn is concentrated — by contract type, tenure, service usage, and payment method.
The end result is a single dashboard that lets you see churn patterns at a glance instead of digging through raw rows.
The workflow follows a standard analyst process:
Raw Dataset → Data Preparation → Pivot Table Analysis → KPI Development → Visualization → Interactive Dashboard
Work involved:
- Reviewing the raw dataset structure (kept on a separate
RAW DATAsheet) - Structuring the data so it could be summarized cleanly through pivot tables
- Building nine separate pivot tables, each covering a specific analytical angle (contract type, tenure group, internet service, payment method, customer segment, service adoption, and number of services vs. churn)
- Calculating KPIs directly from the pivot output (churn rate, average tenure, average monthly charges, total revenue)
- Turning the pivot tables into charts and combining them into one dashboard view
The dashboard is built to answer:
- How much of the customer base has churned, and what does the overall churn rate look like?
- Does contract type affect how likely a customer is to leave?
- Is churn concentrated among newer customers, or spread evenly across tenure groups?
- Does internet service type (DSL, fiber optic, none) relate to churn?
- Does payment method correlate with churn?
- Does the number of services a customer subscribes to relate to how likely they are to churn?
| KPI | Value | What it represents |
|---|---|---|
| Total Customers | 7,043 | Full customer base in the dataset |
| Churned Customers | 1,869 | Customers who left |
| Churn Rate | 26.54% | Share of the customer base that churned |
| Total Revenue | $16.06M | Combined revenue across all customers |
| Avg Monthly Charges | $64.76 | Average amount billed per customer per month |
| Avg Tenure | 32.37 months | Average length of time customers stay before churn/analysis cutoff |
Churn by contract type. Month-to-month, one-year, and two-year contracts are broken out separately, showing churn counts on each.
Churn by tenure group. Customers are grouped into 0-1 year, 1-2 years, 2-4 years, and 4+ years, comparing retained vs. churned counts across each bracket.
Churn by internet service. DSL, fiber optic, and no-internet-service customers are compared for churn counts.
Churn by payment method. Electronic check, credit card, bank transfer, and mailed check are compared for churn counts.
Customer segment split. Customers are grouped into Basic, Premium, and Standard segments, shown as a percentage breakdown.
Service adoption. Adoption rate (% of customers) for individual services — phone service, internet service, online security, online backup, device protection, tech support, streaming TV, and streaming movies.
Number of services vs. churn rate. Customers are grouped by how many total services they subscribe to (1 through 9), with churn rate plotted against that count.
Contract length is one of the strongest retention signals in the data. Month-to-month customers churn at roughly 43% (1,655 of 3,875), one-year contract customers drop to about 11% (166 of 1,473), and two-year contract customers fall to under 3% (48 of 1,695). Customers on longer-term contracts show substantially lower churn, making contract type one of the strongest retention signals in the dataset.
Newer customers are the highest churn risk. In the 0-1 year tenure group, close to 47% of customers churned (1,037 of 2,186), compared to under 10% in the 4+ years group (213 of 2,239). Whatever is driving people to leave, it's happening early — retention efforts likely have the most impact in the first year.
Fiber optic customers churn more than DSL or no-internet customers. Fiber optic churn sits around 42% (1,297 of 3,096), versus about 19% for DSL (459 of 2,421) and roughly 7% for customers with no internet service (113 of 1,526). This is worth digging into further — it could point to pricing, service reliability, or competition in fiber markets, but the dashboard alone doesn't say which.
Electronic check users churn at a noticeably higher rate than other payment methods. Around 45% of electronic check customers churned (1,071 of 2,365), compared to roughly 18-20% for credit card, bank transfer, and mailed check. Payment method looks like it's proxying for something else about the customer (convenience, engagement, or account type), rather than being a direct cause.
Churn risk isn't linear with number of services — it peaks in the middle. Churn rate climbs from about 11% at 1 service to a peak of nearly 45% at 3 services, then steadily declines down to about 5% at 9 services. Customers with just one service and customers with almost every service both stick around; the risk is concentrated among customers with a partial, moderate bundle.
- Microsoft Excel
- Kaggle telecom customer dataset (data source)
- Data preparation and structuring for pivot analysis
- Pivot Tables (9 separate pivot analyses)
- KPI calculation
- Data aggregation
- Chart-based data visualization
- Interactive dashboard design
- Churn and retention analysis
Open Telecom_Churn_Retention_Dashboard.xlsx to see the full pivot table analysis behind this dashboard.
Telecom_Churn_Retention_Dashboard.xlsx— Excel workbook containing the raw dataset, prepared analysis, nine Pivot Tables, KPIs, and interactive dashboarddashboard-preview.png— dashboard preview image
This project was mostly about learning how to move from a raw dataset to something someone can actually act on without opening every tab. Specific things it forced me to work through:
- Structuring a flat dataset so it's usable across multiple pivot tables without redoing work each time
- Deciding which cuts of the data (contract, tenure, internet service, payment method, service count) actually tell a useful story versus which are just noise
- Turning pivot table output into KPIs that mean something on their own, not just raw numbers
- Building a dashboard layout that reads in the right order — overview first, then drill-down detail
- Separating "here's what the chart shows" from "here's what this actually means for the business," which turned out to be the harder half of the project
These are not implemented — just realistic next steps:
- Deeper segmentation combining tenure, contract, and payment method together instead of viewing them separately
- More detailed retention analysis (e.g., cohort-based tenure tracking)
- Automated data refresh instead of a static snapshot
- Rebuilding the analysis layer in SQL
- Migrating the dashboard to Power BI
- A predictive churn model to flag at-risk customers before they leave