Build a Churn Report in Power BI

Analyze telecom churn in Power BI and create testable retention actions.

Introduction

30 Second Summary

A rise in cancelled subscriptions can trigger a rush of guesses about what went wrong. Raw customer totals can make the biggest group look like the biggest problem, even when its individual risk tells a different story.

In this project, you will use Power BI Desktop to build an interactive report that reveals where customer churn concentrates. You will turn those patterns into three retention recommendations with measurable tests.

What You'll Build

You'll present a two-page report that guides stakeholders from high-risk customer segments to testable retention actions.

By the end of this project, you'll have:

An interactive overview where contract-level risk sits beside the customer volume that can otherwise mislead.
A driver analysis that ranks churn patterns with Key influencers before cross-checking them in a contract-by-tenure Matrix.
A retention brief with three testable actions that you can share as a portable PDF.
Secret Mission: Add an AI-guided decomposition tree that follows the highest-churn path across multiple customer dimensions.

Are there any prerequisites?

You'll need a Windows device that meets Power BI Desktop's minimum requirements. No prior Power BI experience is required.

Before We Start

A churn report can guide decisions only when its audience understands the limits of observational evidence. This checkpoint commits you to building a Power BI decision tool that presents retention recommendations as hypotheses to test.

Set Up Power BI and Load the Churn Data

A trustworthy churn report starts with a supported analysis tool. It also needs a traceable source before any chart can carry weight.

In this step, you will install Power BI Desktop before connecting to the public IBM CSV. Power Query Editor then lets you inspect the source before the report uses it.

In this step, get ready to:

Confirm that your device meets the Power BI Desktop requirements.
Install the current Power BI Desktop release.
Load the IBM churn CSV into Power Query Editor.

Check your device and install Power BI Desktop

Power BI Desktop relies on a compatible Windows environment. Checking the operating system and available memory now keeps device limitations separate from later data issues.

Installed memory shows the physical RAM in your device. Available memory shows how much RAM Power BI Desktop can use right now.

Press the Windows key to open Search.
Type Settings and press Enter to open Windows Settings.
Select System in the left sidebar.
Select About.
Confirm that Windows specifications shows Windows 10 or later.
Confirm that System type shows a 64-bit operating system.
Confirm that Installed RAM shows at least 2 GB.

The About page now confirms the operating system architecture and installed memory.

The available-memory figure is easy to miss. Task Manager shows the amount that remains after Windows and other apps take their share.

Press Ctrl+Shift+Esc to open Task Manager.
Select Performance.
Select Memory.
Confirm that Available shows at least 2 GB.
Return to Windows Settings.
Select Display under System.
Confirm that Display resolution is at least 1440x900.

What counts as a supported display?

Power BI Desktop requires at least 1440x900. Microsoft lists 1600x900 (16:9) for a widescreen display.

Press the Windows key to open Search.
Type Microsoft Edge and press Enter to open it.

Microsoft Edge opens successfully. Your device now meets the browser check required by Power BI Desktop.

The Microsoft Store manages automatic app updates. Your next path depends on whether Power BI Desktop is current, outdated, or missing.

Press the Windows key to open Search.
Type Power BI Desktop.
Select the matching tab below based on the result.

A Store installation can take a few minutes. A quiet progress screen during the download is normal.

✔️ I see version 2.157.1354.0 or higher

Select Power BI Desktop from the search results.
Select Help from the top menu.
Select About.
Confirm that the installed version is 2.157.1354.0 or higher.
Close the About window.

Press the Windows key to open Search.
Type Microsoft Store and press Enter.
Search for Power BI Desktop.
Select the official Microsoft listing.
Complete any available Store update.

The Microsoft Store now manages your current Power BI Desktop installation.

Return to Power BI Desktop.

ⓧ I see an older version

Close Power BI Desktop.
Press the Windows key to open Search.
Type Microsoft Store and press Enter.
Search for Power BI Desktop.
Select the official Microsoft listing.
Complete the available update.

The Store replaces the older release with the current version.

Return to Windows Search.
Type Power BI Desktop and press Enter.
Select Help from the top menu.
Select About.
Confirm that the installed version is 2.157.1354.0 or higher.
Close the About window.

ⓧ Power BI Desktop is not installed

Press the Windows key to open Search.
Type Microsoft Store and press Enter.
Search for Power BI Desktop.
Select the official Microsoft listing.
Select Install.

The Store downloads Power BI Desktop and adds it to your Windows apps.

Return to Windows Search after the installation finishes.
Type Power BI Desktop and press Enter.
Select Help from the top menu.
Select About.
Confirm that the installed version is 2.157.1354.0 or higher.
Close the About window.

Good progress. Power BI Desktop is current and ready to load the churn source.

Having trouble with the installation?

Recheck the Windows architecture if the Store says the app is incompatible.
Confirm that your internet connection is active if the download does not progress.
Restart Power BI Desktop after an update if the About window still shows the older version.
Help me troubleshoot my Power BI Desktop installation.

Connect to the IBM churn CSV

The Web connector keeps the source address with the query. This gives your analysis a source that another analyst can trace.

Power BI may show a credential prompt when this device connects to the address for the first time. Follow the prompt for a public web file.

Select Home in the open Power BI Desktop window.
Select Get data.
Select Web.
Paste https://raw.githubusercontent.com/IBM/telco-customer-churn-on-icp4d/master/data/Telco-Customer-Churn.csv into the web address field.
Select Connect.
Follow any credential prompt for the public file.
Select Transform Data when the data preview is ready.

Power Query Editor opens with the IBM Telco Customer Churn dataset displayed as a table.

Why use the Web connector?

The connection records the public IBM address as the query source. Anyone reviewing the report can identify where the customer records came from.

Data preview not loading?

Check that the pasted address ends with Telco-Customer-Churn.csv.
Confirm that Microsoft Edge can open another public website.
Retry the connection after restoring internet access.
Help me troubleshoot the IBM CSV connection.

Prepare the Telco Churn query

A descriptive query name makes the data model easier to follow. Reviewing the column types also catches assumptions before they affect a visual.

Right-click the current query in the left query pane.
Select Rename.
Enter Telco Churn and press Enter.

The query now has the same name that the later measures and visuals use.

The source contains blank-string values in TotalCharges. Keeping that column as text preserves those source values during this project.

Review the type icon in each column header.
Confirm that customerID, Contract, PaymentMethod, and Churn use text types.
Confirm that tenure and MonthlyCharges use numeric types.
Select the TotalCharges column.
Choose the text type if the column shows a different type.
Leave every other source field unchanged.

Before the final check, do you expect this source to include the fields needed for contract-level churn analysis?

Scan the preview headers from left to right.

You should see fields including customerID, tenure, Contract, PaymentMethod, MonthlyCharges, and Churn in the preview.

Missing a source field?

Confirm that the query source uses the exact IBM raw CSV address from the previous substep.
Scroll horizontally if later columns such as MonthlyCharges or Churn are outside the visible preview.
Remove any accidental column-removal step from the query settings.
Help me restore the IBM source columns.

✔️ Awesome, I've got everything!

That is the foundation in place. Keep Power Query Editor open for the next step.

ⓧ I'd like to double check the full setup

Compare your connection and query with these final states.

The web data connection points to https://raw.githubusercontent.com/IBM/telco-customer-churn-on-icp4d/master/data/Telco-Customer-Churn.csv.
The Power Query query is named Telco Churn.
The query contains the IBM Telco Customer Churn source columns unchanged.
The TotalCharges column is typed as text because the source contains blank-string values.

Your source is connected and ready for analysis. Next, you will turn these customer records into a first visual that exposes the limits of raw churn counts.

Build the First Count View

Your Telco Churn query now preserves the IBM source fields you need. The next goal is a report page that shows customer volume across contract types.

Power BI can turn those rows into a Card plus a clustered bar chart. You will inspect the first story those counts tell before deciding how much confidence it deserves.

In this step, get ready to:

Apply the query to the report model.
Build count visuals on the Churn Overview page.
Document the analytical limitation of raw counts.

Create the overview page

Power Query Editor holds your reviewed query changes. Close & Apply loads those changes into the local report model.

Click Close & Apply in the Home ribbon.

Power BI loads the Telco Churn table. You return to a blank canvas in Report view.

The report page tab sits along the bottom of the canvas. This tab controls the page name that readers see.

Double-click the report page tab at the bottom of the canvas.
Type Churn Overview into the page-name field.
Press Enter to apply the new page name.

A Card gives the audience one headline number. Your first Card establishes the total number of customer records behind every later comparison.

Click a blank area near the top of the report canvas.
Select Card from the visual gallery.
Drag customerID from the Telco Churn table into the Card field well.
Open the aggregation menu beside customerID.
Select count as the aggregation.

The Card should show 7043. This confirms that all customer rows are available to the report.

Card Not Showing 7043?

Confirm that the Card uses customerID from the Telco Churn table.
Confirm that the aggregation is set to count.
Help me troubleshoot the customer count in my Power BI Card.

Build the contract count chart

A clustered bar chart compares customer counts across contract categories. A legend separates each category by its Churn outcome.

Click a blank area beneath the Card.
Select Clustered bar chart from the visual gallery.
Drag Contract into the category field well.
Drag customerID into the value field well.
Open the aggregation menu beside customerID.
Select count as the aggregation.
Drag Churn into the legend field well.

You should now see three contract categories. Each category contains separate Yes and No bars.

Before you inspect the bar lengths, predict which contract category will occupy the most horizontal space.

Identify the contract category with the longest combined bar lengths.

You will find that Month-to-month occupies the most horizontal space. The view still cannot compare churn probability within each contract group.

Why Do Counts Fall Short?

A count shows how many customers or churn events appear in a segment. A large segment can produce a long bar because it contains more customers.

The chart does not account for the size of each contract group. The longest bar cannot support a claim that an individual customer faces the highest churn risk.

Document the count limitation

A short annotation keeps the chart's limitation visible beside the evidence. It prevents a stakeholder from treating customer volume as churn probability.

Select Insert in the ribbon.
Select Text box.
Type counts describe volume, not the probability that a customer in that segment will churn. into the text box.
Position the text box beneath the contract chart.

Before the final check, predict whether all three contract categories will each split into two churn outcomes.

Review the Card to confirm that it shows 7043.
Confirm that the chart displays Month-to-month, One year, and Two year.
Confirm that each contract category is split by Churn values.
Read the text box to confirm that the warning appears on the same report page.

The Card shows the full customer count. The chart shows all three contract categories.

Each category is split into Yes plus No outcomes. The warning sits on the Churn Overview page.

That is your baseline working: the report now exposes the exact count-based view that can tempt an audience into an unfair risk comparison.

Your first report page now shows where customer volume concentrates. Next, you will turn this baseline into a fair comparison of churn risk.

Create Comparable Churn Rates

Your first chart exposed a problem with raw customer counts. A large contract segment can produce a long churn bar even when its customers do not face the highest risk.

You will use DAX measures in Power BI Desktop to calculate churn on a comparable percentage scale. Each contract category can then reveal its share of customers who churned.

In this step, get ready to:

Create reusable measures for customer volume and churn risk.
Replace the contract chart's raw counts with a comparable churn rate.
Build a KPI summary for the report page.

Create reusable churn measures

A measure calculates a result from the data currently included in a visual. This active slice of data is called filter context.

Select the Telco Churn table in the Data pane.
Select Home in the ribbon.
Select New measure.
Create the Customers measure by pasting this formula into the formula bar:

Customers = COUNTROWS('Telco Churn')

What Does This Measure Do?

The COUNTROWS function counts the rows available from Telco Churn.
Each source row represents one customer. The result therefore gives you the customer population inside the current filter context.

Press Enter to create the measure.

You should see Customers beneath the Telco Churn table in the Data pane.

Customers Measure Not Appearing?

Check that the formula begins with Customers =.
Check that the table name appears as 'Telco Churn'.
Ask for help with the measure formula.

Select New measure from the Home ribbon.
Create the Churned Customers measure by pasting this formula into the formula bar:

Churned Customers =
CALCULATE(
    [Customers],
    'Telco Churn'[Churn] = "Yes"
)

How Does the Churn Filter Work?

The CALCULATE function evaluates the existing Customers measure inside a modified filter context.
The filter keeps rows where Churn equals Yes.
Reusing Customers keeps the customer-counting logic in one measure.

Press Enter to create the measure.

You should now see Churned Customers beneath the Telco Churn table.

Churned Customers Measure Not Working?

Confirm that Customers already exists before creating this measure.
Check that the churn value uses the exact capitalization Yes.
Ask for help with the churn filter.

Select New measure from the Home ribbon.
Create the Churn Rate measure by pasting this formula into the formula bar:

Churn Rate = DIVIDE([Churned Customers], [Customers])

Why Use DIVIDE?

The DIVIDE function divides churned customers by all customers in the current filter context.
The result becomes the share of customers who churned inside each segment.
The function returns a blank result by default when the denominator is zero. This prevents a division error.

Press Enter to create the measure.

You should see all three measures beneath the Telco Churn table.

Churn Rate Measure Not Appearing?

Confirm that both referenced measures already appear beneath Telco Churn.
Check that each measure name inside the formula uses matching square brackets.
Ask for help with the rate calculation.

✔️ Awesome, I've got everything!

All three measures now appear beneath the Telco Churn table.

ⓧ I'd like to double check the full code

The complete measure set should match this reference.

Customers = COUNTROWS('Telco Churn')

Churned Customers =
CALCULATE(
    [Customers],
    'Telco Churn'[Churn] = "Yes"
)

Churn Rate = DIVIDE([Churned Customers], [Customers])

Replace counts with churn rate

The contract chart needs one comparable value for each category. Percentage formatting makes each bar readable as a share of customers in that contract group.

Select Churn Rate in the Data pane.
Set its format to Percentage in the ribbon.
Select the existing Clustered bar chart on Churn Overview.
Remove the count of customerID from the value field.
Remove Churn from the legend field.
Add Churn Rate to the value field.
Sort Churn Rate from highest to lowest using the chart's sort menu.
Turn on data labels using the visual's format controls.

Before you compare the finished bars, which contract category do you expect to have the highest churn rate?

Compare the three percentage bars in the updated chart.

You should see Month-to-month at the top with the highest churn rate. The chart now compares risk within each contract category.

Chart Still Showing Customer Counts?

Check that the value field contains Churn Rate.
Remove any remaining count of customerID from the chart.
Ask for help updating the visual.

Build the KPI summary

The report needs context beside the contract comparison. A compact KPI area shows the population size before presenting the churn total and rate.

Select the existing Card visual on Churn Overview.
Remove the count of customerID from the Card.
Add Customers to the Card.
Add Churned Customers to the Card.
Add Churn Rate to the Card.

Before the final check, what evidence should tell you that the report now distinguishes customer volume from churn risk?

Confirm that the KPI Card area shows Customers.
Confirm that the KPI Card area shows Churned Customers.
Confirm that the KPI Card area shows Churn Rate as a percentage.
Confirm that the contract chart uses Churn Rate as its value.
Confirm that the contract chart is sorted from highest to lowest.
Confirm that the count-versus-risk warning remains visible.

You should see Month-to-month as the highest-rate contract category. That is the key comparison your original count chart could not make fairly.

Well done. Your report now separates the size of each customer group from the probability of churn inside that group.

KPI Values Missing?

Check that you selected the existing Card before adding each measure.
Confirm that all three fields use measure icons beneath the Telco Churn table.
Ask for help with the KPI area.

Your overview now identifies contract-level risk on a common scale. Next, you will test whether that pattern holds across tenure, payment method, internet service, and technical support.

Triangulate Churn Drivers

Your Power BI Desktop report now compares contract categories using Churn Rate. Month-to-month leads that comparison.

A single-variable chart can hide overlapping customer characteristics. Key influencers ranks candidate factors. A Matrix checks contract risk across tenure bands.

In this step, get ready to:

Build a Key influencers view that ranks candidate churn factors.
Compare contract churn rates across 12-month tenure bins.
Cross-check the ranking with direct rate charts and a TechSupport slicer.

Create the driver ranking

Key influencers evaluates candidate fields against an outcome. It ranks the patterns that most distinguish the selected outcome.

Select the plus button beside the report page tabs.
Double-click the new page tab.
Type Driver Analysis.
Press Enter to save the page name.

You now have a blank page dedicated to testing churn-driver hypotheses.

Select an empty area of the Driver Analysis canvas.
Select Key influencers from the visual gallery.
Drag Churn from the Data pane into Analyze.

The visual now frames churn as the outcome that its candidate factors must explain.

Drag Contract into Explain by.
Drag tenure into Explain by.

The visual begins evaluating contract and tenure as possible explanations for churn.

Drag PaymentMethod into Explain by.
Drag InternetService into Explain by.

The ranking recalculates as payment and internet-service fields join the candidate set.

Drag OnlineSecurity into Explain by.
Drag TechSupport into Explain by.

The visual now tests whether support-service categories help distinguish churn outcomes.

Drag MonthlyCharges into Explain by.

Monthly charges now contribute a numeric candidate to the ranking.

Confirm that Expand by is empty.
Select Yes in the outcome dropdown.

You should see the visual rank factors that distinguish customers whose Churn outcome is Yes.

How Does the Ranking Work?

The Analyze field defines the outcome under investigation.
The Explain by fields define the candidate factors that Power BI compares.
The selected Yes outcome focuses the ranking on customers who churned.
The ranking supports hypotheses about associations. It does not establish causal proof.

Visual Not Ranking Factors?

Check that Churn is inside Analyze.
Confirm that Expand by has no field.
Confirm that every candidate field is inside Explain by.
Help me troubleshoot my Key influencers setup in Power BI Desktop.

Build the tenure interaction matrix

An influencer ranking evaluates candidate patterns across the dataset. A contract-by-tenure interaction check reveals whether contract risk changes across stages of the customer relationship.

Right-click tenure in the Data pane.
Select New group.

You should see the Groups dialog with controls for turning the numeric field into bins.

Enter 12 in Bin size.
Select the confirmation button in the Groups dialog.

You should see a new tenure field with the (bins) suffix in the Data pane.

Select an empty area of the Driver Analysis canvas.
Select Matrix from the visual gallery.

You should see an empty Matrix frame on the page.

Drag Contract into Rows.
Drag the new tenure field with the (bins) suffix into Columns.

The Matrix now lays out contract categories against 12-month tenure bands.

Drag Churn Rate into Values.

You should see percentage values across the Matrix cells. Differences between cells expose contract and tenure combinations with distinct churn rates.

Why Check the Interaction?

A contract category can contain customers at many tenure stages. The Matrix separates those stages so a broad average cannot hide a concentrated high-rate group.

Agreement between the Matrix and the influencer ranking strengthens a hypothesis. A disagreement signals that the pattern needs more investigation.

Cross-check the ranking

A ranked factor becomes more defensible when a direct rate view shows the same pattern. Separate charts also make the evidence easier to explain to a stakeholder.

Select an empty area of the Driver Analysis canvas.
Select Clustered bar chart from the visual gallery.

You should see an empty bar-chart frame beside the existing visuals.

Drag PaymentMethod into the category field well.
Drag Churn Rate into the value field well.

You should see churn rates compared across payment methods.

Select another empty area of the Driver Analysis canvas.
Select Clustered bar chart from the visual gallery.

You should see a second empty bar-chart frame.

Drag InternetService into the category field well.
Drag Churn Rate into the value field well.

You should see churn rates compared across internet-service categories.

Select another empty area of the Driver Analysis canvas.
Select Slicer from the visual gallery.

You should see an empty slicer frame on the page.

Drag TechSupport into the slicer field well.

You should see the available technical-support categories in the slicer.

Before you compare the views, which suspected driver do you think will appear in more than one visual?

Review the highest-ranked factors in the Key influencers visual.
Compare those factors with the two Clustered bar charts.
Compare those factors with the contract-by-tenure Matrix.
Select one category in the TechSupport slicer.

You should see the rate views recalculate for the selected technical-support category. This response shows how filtering can change the apparent strength of a pattern.

Restore the TechSupport slicer to its unfiltered state.
Confirm that the Key influencers question analyzes Churn for the Yes outcome.
Confirm that the Matrix displays different Churn Rate percentages across its cells.

You now have a ranked factor view plus transparent rate comparisons. Together, they support evidence-backed churn hypotheses without presenting associations as proven causes.

Direct Views Not Responding?

Check that both bar charts use Churn Rate as their value.
Check that the Matrix uses Churn Rate in Values.
Confirm that the slicer uses TechSupport.
Help me troubleshoot the visuals on my Driver Analysis page.

That is a meaningful analytical upgrade. Next, you will turn these supported patterns into prioritized retention actions with measurable tests.

Create a Retention Action Brief

Your two analysis pages now show where churn concentrates. The next challenge is turning those patterns into decisions that a mixed business audience can act on.

In Power BI Desktop, you will create a retention brief that links every recommendation to visible evidence. The data supports associations, while a randomized or phased test determines whether an intervention causes improvement.

In this step, get ready to:

Create an executive summary that distinguishes observed association from causal proof.
Prioritize three retention recommendations with owners, metrics, and causal tests.
Package the finished report as local PBIX and PDF files.

Create the executive summary

An executive summary turns the strongest churn rate pattern into a clear business priority. It also sets an evidence boundary so readers do not mistake an observed pattern for a proven cause.

Why causal language matters

The report reveals associations between customer characteristics and churn. A controlled test is still required to show whether a proposed intervention changes the outcome.

This distinction keeps the brief useful without promising more certainty than the data provides.

Return to the Driver Analysis page from earlier.
Use the Matrix to confirm that first-year Month-to-month customers form the strongest observed segment.
Compare that result with the Key influencers ranking.
Click the plus sign beside the report page tabs to add a blank page.
Rename the new page Retention Actions.

You should see a blank report page with Retention Actions selected in the page tab bar.

Select Text box from the Insert ribbon.
Click near the top of the canvas to place the text box.

You should see an empty text box ready for the executive summary.

Start the summary by naming first-year Month-to-month customers as the strongest observed segment.
Add one sentence explaining why this segment matters to the business.

Your summary now connects the strongest visible pattern to a retention priority.

Finish the summary with an explicit caveat that the report shows association instead of causal proof.

Good progress. The new page now gives decision-makers a clear finding with an honest evidence boundary.

Prioritize three retention tests

A useful recommendation connects a high-rate segment to an intervention that someone can own. Its experiment design must also define how the business measures success without creating a new problem.

What each recommendation needs

A target segment identifies the specific customers included in the action.
A proposed intervention describes the retention change being tested.
An owner identifies the team responsible for delivery.
A primary success metric measures the intended retention outcome.
A guardrail metric detects an unwanted side effect.
A randomized or phased test provides a comparison that can support a causal decision.

Return to the Driver Analysis page.
Compare the Key influencers ranking with the payment method bar chart.
Compare the Key influencers ranking with the internet service bar chart.
Use the TechSupport slicer to check whether a candidate segment remains high risk.
Identify two additional segments with visible rate differences and meaningful customer volume.
Return to the Retention Actions page.

Select Text box from the Insert ribbon.
Place the text box below the executive summary.

You should see a second text box with enough space for three recommendations.

Complete Recommendation 1 for first-year Month-to-month customers using the six-part structure above.
Complete Recommendation 2 for the next high-rate segment using the same structure.

You should now see two testable actions with named owners and measurable outcomes.

Complete Recommendation 3 for the remaining high-rate segment using the same structure.
Order the three recommendations from highest to lowest priority.
Compare each recommendation with the visible rate difference that supports it.
Check that each affected segment contains enough customers to justify action.
Remove any recommendation that fails either evidence check.
Confirm every remaining recommendation includes all six required fields.

Unsure which segments to prioritize?

Favor segments that appear consistently across the Key influencers ranking and direct rate views. Reject a striking percentage when the affected customer group is too small to justify action.

Ask for help interpreting the evidence: help me prioritize these churn segments without overstating causality.

Save and export the report

A local .pbix file preserves the editable report. A PDF gives stakeholders a portable copy that does not require Power BI Desktop.

Select File from the top menu.
Select Save.

You should see the local save dialog for the report.

Choose your report folder as the local folder for both artifacts.

The save dialog should now show your chosen Windows folder.

Enter subscription-churn-root-cause.pbix as the file name.
Click Save.

Your editable report is now stored locally as subscription-churn-root-cause.pbix.

What the local file preserves

The saved .pbix keeps the report pages with the data model. It also retains the Telco Churn query, tenure bins, measures, and cached web connection.

The report stays on your device. You do not need Publish or a Power BI service sign-in.

Before you export, predict whether each visible report page will become its own page in the PDF.

The export can take a few minutes while Power BI Desktop prepares every visible page.

Select File from the top menu.
Select Export.
Select Export to PDF.
Name the exported file subscription-churn-root-cause.pdf.
Save the exported file in your report folder.
Wait for the export process to finish.

The exported PDF displays in your default PDF viewer. You should see separate pages containing Churn Overview, Driver Analysis, and Retention Actions.

Missing a page from the PDF?

Return to Power BI Desktop to confirm that all three report page tabs are visible.
Export the report again after making every required page visible.
Check that the PDF filename is subscription-churn-root-cause.pdf.
Get targeted help: help me troubleshoot my Power BI Desktop PDF export.

That is the decision layer complete. Your local report now moves from churn evidence to three testable retention actions. The PDF gives stakeholders a portable copy.

Secret mission

Add an AI-Guided Decomposition Tree

Trace churn risk through several customer characteristics using an AI-guided Decomposition tree. Turn the resulting high-churn segment into a testable retention hypothesis.

Clean Up Your Resources

All project artifacts stay on your Windows device. They create no paid resources or recurring charges.

Choose whether to keep your work, pause by closing Power BI Desktop, or delete the local artifacts.

Resources you used:

The local subscription-churn-root-cause.pbix report file.
The local subscription-churn-root-cause.pdf stakeholder export.
The cached Power BI Desktop web connection for the public IBM CSV at https://raw.githubusercontent.com/IBM/telco-customer-churn-on-icp4d/master/data/Telco-Customer-Churn.csv.

Keep everything running

No action is needed. Choose this option if you want to keep refining the report or sharing the PDF.

Retain subscription-churn-root-cause.pbix in your chosen Windows folder.
Retain subscription-churn-root-cause.pdf in the same folder.
Leave the cached web connection available for future data refreshes.

Pause - I'll come back to this later

Close the local app to stop editing. Your files remain on disk with no ongoing service or cost.

Close Power BI Desktop.
Leave subscription-churn-root-cause.pbix in your chosen Windows folder.
Leave subscription-churn-root-cause.pdf in the same folder.
Resume later by opening subscription-churn-root-cause.pbix in Power BI Desktop.

Delete - I don't want to use this again

Deleting these local artifacts is final on this device. The public IBM CSV remains available if you decide to rebuild the report.

The cached web connection lives inside Power BI Desktop.

In Power BI Desktop, select File.
Select Options and settings.
Select Data source settings.
Select the cached source for https://raw.githubusercontent.com/IBM/telco-customer-churn-on-icp4d/master/data/Telco-Customer-Churn.csv.
Clear the selected source.
Confirm that the clear action completes.
Close Power BI Desktop.

Your report file and PDF remain in the Windows folder you chose.

Use your Windows file browser to return to the folder where you saved subscription-churn-root-cause.pbix.
Select subscription-churn-root-cause.pbix.
Hold Ctrl while selecting subscription-churn-root-cause.pdf.
Delete the selected files.
Confirm that neither filename remains in the folder.

Nice Work!

Outstanding work! Your local Power BI report now turns churn data into testable retention decisions for a mixed business audience.

What you learned:

Built an interactive churn report from the public IBM dataset. Exposed why customer counts measure volume. Replaced them with churn rates for fair segment comparisons.
Created reusable DAX measures for customers, churned customers, and churn rate. Used Key influencers to rank churn patterns. Cross-checked the ranking with a contract-by-tenure Matrix. Compared payment method, internet service, and technical support directly.
Turned observed associations into three retention recommendations. Each action includes an owner, a success metric, a guardrail metric, and a causal test. Saved the reusable report as subscription-churn-root-cause.pbix. Exported the analysis as subscription-churn-root-cause.pdf for stakeholders.
Secret Mission: Added an AI-guided Decomposition tree that follows a multi-level High Value path to a higher-churn segment. Framed the combined segment as a hypothesis with a validating experiment.

Ready to quiz yourself?