Banking & Finance

Insights Study- Predicting insurance renewal with machine learning

How we use data science and CX data to predict insurance renewal

AI is set to disrupt all industries and the insurance sector is no exception. We are entering a new era where companies are analyzing a huge volume of disparate data in real-time to predict customer behavior and improve financial outcomes.

In this insights study, we’ll explain how NumrCXM uses customer satisfaction scores and machine learning to predict insurance renewal.

Predictive Analytics

Predictive analytics encompasses a wide variety of statistical techniques that analyses historical and current data to make predictions about the future. 

Historical data is then fed into an algorithm (that considers the key trends and patterns) and a Model is developed. This Model is then applied to the current data sets to predict what will happen in the future.

Predictive Analytics

Financial Benefits of Predicting Insurance Renewal

Predicting customers who are most likely to default on their premium payments gives you the power to target and retain them.

Following up with only at-risk customers,

·      Reduces churn

·      Increases renewal

·      Improves revenue, and

·      Saves time and resources

Understanding why customers are churning empowers you to fix those issues and prevent the possibility of a large-scale churn.

How Numr Predicts Insurance Renewal

We use machine learning and experience data to predict customers who will not pay premiums.

STEP 1- Data Pre-Processing

We start by preparing the dataset. This starts with asking the client for historical customer data. This data includes

·      NPS scores

·      Transactional information

·      Demographic data, and more

Data preparation usually comprises of 3 main steps- encoding, imputation, and merging.

Encoding data includes turning categorical variables into binary so that the model can read it.

In real-life, data is rarely perfect. There are usually some missing data that can render entire rows useless. Imputation rescues those rows. We impute values to fill the blank cells.

The last step is merging the encoded data with numerical data.

Data Preparation

Data preparation is now complete.

STEP 2- Split Data into Train and Test

To evaluate the performance of the renewal model, we split the dataset into train and test. This means we use 70%of the data to train the model and the remaining 30% to test it.

STEP 3- Develop the Model

We use the train dataset to then develop the model. In this case, we use a classification model to find customers who will not renew their insurance.

Instead of merely relying on Accuracy, we look at other scores such as the F1 and Precision Score.

STEP 4- Test the Model

Once the model has been developed, it is ready to be tested. We use the remaining 30% dataset to compare the predicted outcome with the actual outcome.

After we are satisfied with the accuracy of the predictions, the model is ready to be deployed.

It will predict customers who are most likely to churn and not renew their insurance. Over time, we keep feeding the model with more data to improve it.


Predicting insurance renewal allows companies to streamline their sales efforts. If you know which customers are going to churn a long time before they actually do, you can target them and find out why. More often than not, you can retain them by proactively solving their issues.

Talk to a Specialist

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form. Please try again.

You are opting-in to Numr's marketing communications. We use the information you provide to us to contact you about our relevant content, products, and services. We value your privacy and you can learn all about our data collection practices in our privacy policy.