How machine learning personalisation increases payments by 7%

Personalisation has fast become the norm. Now, 71% of consumers now expect companies to create personalised interactions and on the flip side, 76% feel frustrated when their experiences aren’t bespoke. 

The business benefits of personalisation are also significant. McKinsey & Company found that it can: 

• Reduce customer acquisition costs by up to 50%
• Lift revenues by 5 -15%
• Increase marketing ROI by 10 - 30%

So how do we apply this in debt collection? Introducing, the message recommender.

Personalise every single message

Collect, our intelligent debt collection product, is powered by machine learning models. Each of these models enhance a particular point of the collections experience. For example, our message scheduler model ensures that emails land in a customer’s inbox at the time they’re most likely to take action. 

The message recommender is our latest machine learning model, that personalises every single message that every single customer receives.

Intuitive to where a customer is on their journey 

When a customer is referred to Collect, their profile and behaviour is assessed to determine how they’re most likely to engage with their debt. The message recommender model selects a message that has the right tone, content and call to action for an individual customer, so they’re most likely to feel empowered and take action. 

To make an informed selection, the model measures the performance of every single message that’s sent by Collect. It analyses:

• Open rate - the rate at which the message was opened
• Click rate - the rate at which the link in the message was clicked
• Conversion rate - the proportion of customers who resolved their debt via the message (by setting up a payment plan or paying in full)
• Number of times the message has been used 

To create a truly intuitive customer experience, the model combines this information with:

• Where the customer is on their collections journey - for example, have they already received two emails from us, or will this be their first message?
• The location of the customer - to ensure the message meets all the compliance requirements for that region

Once the model has analysed all the above, it narrows down the best potential messages to send. Finally, it looks at the past performance of all the remaining messages and selects the highest performing one. 

A key aspect of personalisation is ensuring that the entire experience is individualised and intuitive, which is why the message recommender model is used to optimise every communication a customer receives.

Increase in conversion and click rates 

The latest version of the message recommender model was launched in March 2023. When compared to those who have been sent communications at random, customers that receive messages that have been recommended by the message recommender have a:

• 6.9% increased conversion rate
• 5.4% increased click rate

These significant uplifts are noted across all customers, regardless of where they are in their collections journey. But, the real icing on the cake? Machine learning models are exactly that - learning. They become smarter, more precise and produce improved recommendations with every interaction. Find out more here in our guide to intelligent debt collection here.