In this article, we will explore several artificial intelligence features in Power BI applied to a telecommunications company. Key Influencers analysis will help us identify the main factors driving churn, highlighting the impact of high daytime charges, the number of support calls, and having international plans. Next, we will use the forecasting feature to predict future sales based on time series, supporting stock management, marketing planning, and risk mitigation. Anomaly detection will allow us to spot unexpected deviations in sales, enabling a fast strategic response. As described in the previous article, we used the Power BI Design Kit to ensure visual consistency.
The Telecom Case
Analyzing Churn with Key Influencers
Churn rate is one of the most important KPIs in industries such as telecom, gyms, insurance, and other subscription-based services. Power BI offers a feature called Key Influencers that uses machine learning to uncover what drives churn.
How does it work?
- Power BI analyzes patterns in your data and detects the factors influencing customer churn
- The visual highlights what increases or decreases churn probability
- You can click each factor to see how it impacts the result
What drives churn the most?
- High daytime charges – Customers spending more than €48.50 on daytime calls are 6.21x more likely to leave
- Frequent support calls – Calling support more than 4 times increases churn risk by 4.91x
- Active international plan – These users are 3.77x more likely to churn, possibly due to price comparisons
How to reduce churn?
- Create more competitive packages for high-usage customers
- Improve customer service and reduce waiting time
- Offer incentives to retain premium users
Forecasting Future Sales
If you knew how your sales would perform over the next three months, wouldn’t your decisions be more accurate? Power BI’s forecasting feature does exactly that: it analyzes past trends and projects future values.
How does it work?
- Power BI analyzes historical data and creates a forecast for upcoming periods
- The black line shows the prediction, and the light blue area represents the uncertainty range
What did we learn from the forecast?
- Overall sales trend shows growth, with fluctuations
- Forecast suggests a mild recovery after a dip
- High uncertainty means historical data is highly variable
How to apply it in your business?
- Stock management – avoid over or understocking
- Marketing planning – align campaigns with expected sales behavior
- Risk analysis – support negotiations with banks and investors
Detecting Sales Anomalies
If sales suddenly spike or drop, understanding the reason is crucial. Power BI’s anomaly detector helps you spot these unexpected patterns.
How does it work?
- Power BI defines a range of expected values
- When a data point falls outside this range, it’s flagged as an anomaly
Real-world example
- July 2023 showed an unusual sales spike
- Sales dropped afterwards and recovered in 2024
Possible causes
- Special promotion that boosted performance
- Data entry error or system issue
- External factors – competition or market shifts
Impact on decision-making
- Replicate winning strategies (if it was a positive spike)
- Fix operational issues (if caused by internal problems)
- Monitor future trends to respond faster
Customer Segmentation with Clustering
Not all customers are the same. Clustering helps group them by behavior, enabling more targeted strategies.
How does it work?
- Power BI uses machine learning to cluster customers by:
- Total sales
- Number of purchases
- Purchase frequency
Example segmentation
- Cluster 1: Many customers, low average spend
- Cluster 2: Mid-tier spend, more loyal
- Cluster 3: Premium clients, high ticket value
How to apply it in your business?
- Give discounts to low spenders to boost volume
- Build loyalty programs for mid-tier clients
- Improve service and perks for premium users
Exploring Data with the Decomposition Tree
The Decomposition Tree allows multi-level exploration of metrics like sales, offering a drill-down experience.
How does it work?
- Start with total sales
- Break down by plan, product, customer type, or region
- Identify key drivers of revenue
Real-world insights
- Customers without international plans generate more sales
- VoiceMailPlan is the top-selling product
- Individual clients buy more than business customers
How to apply?
- Adjust marketing for your most profitable segments
- Create targeted campaigns for underperforming regions
Turning Data into Text with Smart Narrative
Power BI can automatically generate descriptive summaries of your data, helping you present insights with clarity.
Example insights generated:
- “Individual customers grew by 14%, while business customers declined by 8%.”
- “Bundle category grew 98% between April and December 2024.”
Why use it?
- Saves time – Automatic summaries of complex analysis
- Supports decisions – Clear and actionable insights
- Fully customizable – Edit text to fit your context
Which of these features did you find most useful? Leave your thoughts in the comments! 👇🚀
