If you are a product manager or marketing professional who wants to understand what drives certain customer behavior, you can find that out with a few clicks.
What drives your customers to churn? What leads to a wow experience? What makes them buy your product? In a nut-shell, understand the factors that drive a metric you’re interested in. Power BI makes excellent use of Machine Learning and AI capabilities to analyze your data, rank the factors that matter, and display them as key influencers. It also creates smart segments of the customers based upon the combination of factors.
The good part is- you don’t need to have expertise in data science. All you need to do is upload or connect your raw data and make a few clicks. Let’s take a sample dataset of telcom Customer Churn available on Kaggle-https://www.kaggle.com/blastchar/telco-customer-churn. The target variable is churn i.e. customers who left within the last month. The data set includes information about:
- Customers who left within the last month — the column is called Churn
- Services that each customer has signed up for — phone, multiple lines, internet, online security, online backup, device protection, tech support, and streaming TV and movies
- Customer account information — how long they’ve been a customer, contract, payment method, paperless billing, monthly charges, and total charges
- Demographic info about customers — gender, age range, and if they have partners and dependents
Let us try to understand what factors drive churn.
Step 1: Load data in PowerBI
File -> Get Data -> Text/CSV
There are 7043 rows and 19 columns with customer attributes which we will use as input factors with last column ‘Churn’ as the output variable.
Step 2: Drag & Drop
Click on Key Influencers under Insert section. Drag target variable (churn here) under ‘Analyze’ and all factors you want to keep under ‘Explain By’. I have dragged all 19 variables as factors in the image below.
Step 3 : Read results
3.1 Understanding Key Influencers
You will observe that the results automatically update as you add or remove the variables from the list. Let’s take a look at the key influencers. As we can see in the image, having Month-to-Month contract is the top factor that contributes to the Churn. Precisely, Month-to-Month contract customers are 6.32 more likely to churn as compared to other contract types (One Year, Two Year).
Second most important factor is OnlineSecurity, followed by TechSupport where 3.63 and 3.51 are the respective lifts in churn.
So far, you’ve seen different categorical variables. ‘Contract’ is a categorical variable with three labels: Month-to-month, One Year and Two Year. Power BI shows additional details for the selected variable in the right pane where you can see the comparative effect of each label on the likelihood to churn. The dashed line shows Avg. Churn % of all values except for the selected key factor (Contract is Month-to-Month).
Let’s look at details when the variable is continuous — eg. TotalCharges. The fourth most important factor is ‘TotalCharges is 68.45–96.45’. Notice that PowerBI has automatically binned the continuous variable. Distribution in the right pane shows that the relationship between the TotalCharges and Churn is not linear. I find this optimal binning to be super useful. As a retention manager, you can give your special focus on the customer segment with 68.45–96.45 as TotalCharges.
Till now, we explored Key influencers tab to assess each factor individually and ranked them. But in reality, it is the combination of factors that affect the metric. Let us explore this awesome functionality of PowerBI where it makes segments by grouping key influencers.
3.2 Extracting user segments
Click Top segments tab to see how a combination of factors affects the churn. The segments are ranked based on the percentage of records where the condition is met. The size of each segment bubble represents how many records (population count) are in the segment.
In our case, 5 segments are discovered. These segments are ranked by the percentage of ‘churn’ within the segment. Segment 1, for example, has 75.9% customers who churned. The higher the bubble, the higher the proportion of low ratings. Selecting a bubble displays the details of that segment. Let us understand Segment 1.
As we can see, Segment 1 users are ‘Month-to-Month’ subscribers, have Fiber optic, and are new subscribers (tenure 4 months or less). In this segment, 75.9% of the customers churned which is 49 percentage points higher than average. Segment 1 also contains approximately 7.2% of the data, so it represents an addressable portion of the population.
You can further drill down this segment by clicking on “Learn more about this segment” and see what other factors influence this segment. Similarly, we can understand the other 4 segments created.
I found this PowerBI tool to be very handy and quick. It is fully interactive, meaning you can use filters, slicers, and selections on other visuals to affect the results. It certainly enables product and marketing strategists to leverage ML/AI without technical know-how. In my org, I made an automated PowerBI dashboard with A/B test experiments data. Key Driver Analysis was an essential part of it. It helps Product and Marketing managers understanding what drives their experiment success or failure and also helps in optimizing future experiments.
Needless to say- As a Data Scientist, I still like to play with my raw data with different approaches- Decision Trees, Classification algorithms etc. Key Driver Analysis is not about a single technique. If you are interested in knowing how PowerBI does this, here is the link I found-