It’s a tale as old as time — why am I losing clients? The maniacal focus on losing clients is really driven by customer acquisition cost (CAC), the dreaded cost of landing new customers. Churn analysis is more important than ever, as the cost to retain existing customers is almost always exponentially lower than the cost of acquiring new ones.
According to Hubspot, cost of acquisition is increasing because:
1. Consumer trust in business is eroding
2. Marketing costs are rising
3. Increased competition (at the source — from marketers themselves like Google with competing offerings)
4. Regulatory environment is more complex
So how can you identify customers who are most likely to churn and come to the rescue? By keeping them as customers and avoiding having to replace a customer with a high CAC, you are directly contributing to profit preservation.
Today, churn analytics requires several tools: a data prep tool, a dashboarding/EDA tool, and a modeling tool such as Python or potentially SAS. What if you’re just one person that needs to accomplish this? What if you don’t want to wait for a data scientist to help you, or a report developer to create the view you need?
I want to show you how you can accomplish churn prevention in one tool, without needing technical skills.
For this blog, we will use this credit card churn dataset available from Kaggle. Loading data is straightforward and simple. Once loaded, you can do some data prep right within Tellius. Here we did some simple type changes via a point and click interface. From type changes to complex aggregations, you don’t need to be a SQL expert to manipulate your data.
Point And Click Data Prep Example
Now to explore and understand our dataset. You can use natural language search to visually inspect the data you are working with. Let’s take a look at how someone might do that:
Exploring our churn dataset
Looks like we have 16% of our customers who churned. Our field list shows quite a few metrics like credit limit, age, etc. Let’s compare churned customers vs existing customers for credit limit and number of dependents to see if we can get a sense of any major differences:
More Exploratory Data Analysis
Looking at our two different classes visually for all of these metrics would be hard — how do we know which had the biggest impact to churned customers? What about when we think about multivariate analysis, where we want to consider multiple variables to understand if combinations of data? It’s almost impossible to understand what drives churn from visual analysis.
Instead, let’s try to use some automated insights from Tellius. We’ll kick off two:
1. A cohort insight that will identify the largest differences between our two classes of customer
2. A key driver insight, that will identify which features are most important as well as identify segments of customers to target for churn
We type these questions to kick off insights
Kicking off insights can be accomplished in Tellius via search, clicking on visualizations or through a prompt menu. Let’s interpret our cohort insight first:
Cohort Insight: What
First, we are presented with the “What” tab in the Cohort insight. At the top, we can see that the number of Churned Customers are represented with Blue, with Existing Customers with Yellow. Below is a summary that identify where the largest differences in churn counts were observed, and a ranked list of those differences below.
This particular insight shows us that the largest differences in churned vs existing customers were in the Blue card category and married or single individuals.
Cohort Insight: Why
Moving to our “Why” tab exposes interesting combinations of metrics that had the most striking differences between churned and existing. Churned customers had a significantly lower revolving balance, average utilization and total transaction count on their credit card.
What can we surmise from this data? Customers who use their cards less, and more likely to churn and go to another credit card provider.
What do we do?
One assumption we can make is that promotions or offers will entice your current customers to use their card more. Lower interest rate offers, incentives for shopping at particular retailers or temporary cash back offers can help us to increase our current customer’s credit card usage, thus reducing their churn.
However, these offers are expensive. Who should we target for the highest impact? How about our customers who are most likely to churn.
Identifying segments of customers to target is often a univariate exercise: based on only one data point. Multivariate segments can provide more powerful, super targeted segments that we can attack with our offers. Finding these segments often take a data scientist and advanced modeling or machine learning knowledge.
With Tellius, we can use our second insight: a key driver insight.
Key driver insight
This key driver insight shows us the features that drive churned customers first in our feature importance section. This verifies the data we saw in our cohort insight, where transaction count and revolving balance were shown to be important metrics that may indicate customers who will churn.
The segment summary identifies the criteria that make up this segment. For the above, our overall dataset had a 16.1% churn rate, while this segment of customer that met this criteria had a 94.4% churn rate. Targeting this segment with offers will help us to reclaim customers from a segment where only 1 in 20 customers remain with us.
Often, this is where analysis stops due to technical barriers. How can we flag this segment in our data and provide an actionable list of customers to target with offers?
With Tellius, you can automatically apply a flag to your dataset without having to know SQL by using Explore Segment → Smart Insights. From here, we are able to create a detailed list of customers that should be targeted with a promotion, with the goal of turning around our worst performing segment of customers when it comes to churn.
A filtered list ready for action
Churn Analytics is easy with Tellius
Churn analytics has historically been viewed as too technical and too hard for any one person. Legacy tools, multiple teams and communication barriers cause churn analytics to be a slow process with output that is out of date by the time it is ready.
Check out a free trial of Tellius to lower your churn rate, keep your customers and avoid high cost of acquisition of new customers.