Video
Multiple Sources of Data
Let’s start by diving into the data.
Sources:
- Product usage data
- Customer demographics
- Tickets rating data
Schema:
- Customer usage data
- Customer demographics and subscription details
- Performance metrics – Customer Retention Number, Total Amount
- Plan details
- Ticket ratings
Let’s dive into how a churn analytics team can go from connecting to all your complex data sources to delivering meaningful insights to the business on areas to take action.
Preparing the Data
First, we start with loading data from multiple sources, including both internal and external sources.
Once we have the data loaded, Tellius helps us evaluate the quality of the data by inferring the data types of each column and automatically detecting any inconsistencies in the data.
Before analyzing the data, we need to perform feature engineering to extract more information on the ticket details, specifically to generate a churn flag using tenure and last payment date. Tellius Prep enables users to leverage point-and-click data transformations or even existing SQL/Python scripts to create new variables directly in the platform.
We calculate tenure based on the subscription start date and last payment date. We define categories based on the ticket score. We also create a churn flag based on the last payment date.
Step 1: Calculate tenure for a customer using SQL
SQL:
select *, CAST( months_between( last_Payment_date, subscription_start_date ) AS INT )
as Tenure from table_name
Step 2: Create a Retention Flag using SQL
SQL:
select *, (case when months_between(current_date, last_Payment_date) > 3
then "False" else "True" end) Retention_Flag from table_name
Tellius Prep provides full flexibility for advanced users to incorporate the required business logic by manipulating the data in any way or creating Calculated Calculations, as shown below.
Exploratory Analysis
Now that the data is ready for analysis, let’s begin answering ad-hoc business questions using Tellius’s powerful natural language Search interface that understands your question, calculates an answer in seconds, and returns the best fit visualization based on the variables used in your question.
Business Question: What is the monthly customer churn?
SEARCH –> monthly customer churn
Some months experienced more churn than others, including a huge spike in September 2019. Let’s see what the relationship between churn and product is.
Business Question: Monthly churn by product type?
SEARCH –> monthly churn by product type?
Automated Insights
Tellius’s automated insights discovered that one segment of customers is 43% more likely to retain than others when training time is high and the product support package is platinum.
In a matter of minutes, Tellius scanned millions of data points and uncovered the key factors driving customer retention. Unlike manual, hypothesis-driven approaches, Tellius is providing us with important facts that ensure we do not miss out on any opportunities. How? Tellius removes bias from the analysis and automatically diagnoses every possible combination of when customers are retained, churned and highlights the key drivers for this segment. With these new insights, the business quickly pivots its strategy and reprioritizes its field force.
Business Question: How do retained vs churned customers compare?
In a matter of minutes, Tellius analyzed the differences between churned and retained customers and uncovered the fact that retained customers have more sessions, their training time is high, and they use Travel Assistant tool and TRAFFICLIVE products. Looking at churned customers, we see that customers who have low training time and are not satisfied with the ticket resolution churned more frequently. A potential action this insights drives would be to improve our ticket management system.
Predictive Analytics
Now that we have identified high and low retention customers in our current books, the next step is to predict at-risk customers via a churn prediction classification model that identifies the key features that are important to determining if a customer will churn. Tellius’s AutoML capabilities can predict customer behavior and identify risky customers with a single click so businesses can take immediate action on these customers.
Under the hood, Tellius’s AutoML scans customer attributes via common classification models used in predicting churn such as Logistic Regression, Random Forest, and Naive Bayes Algorithm — in a tournament-style approach to find a model with the highest predictive power.
The resulting feature importance reveals that the different usage of the product (HR_usage, Travel_Assistant_Usage, Product_Management_Usage, Business_Reporting_Usage, Analytics_Usage), number of sessions, training time and support package contribute the most towards the target of churning or not.
Actionable Insight
Based on the above trained model, we can predict high-risk customers for the current year.
We see that there are ~16k customers at risk of churning (9.32% of total customers this year) of which 12K customers (75%) are at high risk. Previous insights of the differences between retained and churned customers suggest that there are a few steps to pursue with this population of customers, including targeting this population for more training sessions and moving them from the silver support package to gold or platinum.
Summary
As we showcased above, Tellius delivers value by expediting customer retention analysis and churn prediction to improve business outcomes by understanding the drivers of those behaviors.
Tellius is useful for a variety of other eCommerce & Retail applications or download our Guide to AI-Driven Analytics for eCommerce. Take Tellius for a free 14 day spin (no credit card necessary) today!
